U.S. patent number 6,606,171 [Application Number 08/948,108] was granted by the patent office on 2003-08-12 for digitizing scanner.
This patent grant is currently assigned to Howtek, Inc.. Invention is credited to Mark R. Fernald, Richard F. Lehman, Jeffrey D. Renk, Calvin M. Winey.
United States Patent |
6,606,171 |
Renk , et al. |
August 12, 2003 |
Digitizing scanner
Abstract
A digitizing scanner particularly for scanning transparent films
such as X-rays provides an improved illuminator for transmitting
light through the film. The illuminator defines a line array of a
plurality of individually calibrated and controlled LEDs. The LEDs
are calibrated by determining their relative points of projection
on a CCD camera array. The camera array scans the LEDs and adjusts
them individually to produce a predetermined illumination pattern
from the group. The adjustment occurs over a plurality of cycles
that address cross-talk between LEDs in the array. The camera
includes anti-reflection elements to minimize bounce-back of image
light and noise suppression circuitry to reduce low-level signal
noise. A central processing unit, interconnected with the camera
assembly includes a pixel averager to reduce the inherent
resolution of the CCD to a desired level and to attenuate further
noise. The scanner can include an illuminator for illuminating a
scannable opaque bar-code strip and size-measurement circuitry for
determining the relative size and location of the scanned
image.
Inventors: |
Renk; Jeffrey D. (Derry,
NH), Lehman; Richard F. (Nashua, NH), Fernald; Mark
R. (Amherst, NH), Winey; Calvin M. (Carlisle, MA) |
Assignee: |
Howtek, Inc. (Hudson,
NH)
|
Family
ID: |
27663735 |
Appl.
No.: |
08/948,108 |
Filed: |
October 9, 1997 |
Current U.S.
Class: |
358/475;
358/487 |
Current CPC
Class: |
H04N
1/00002 (20130101); H04N 1/00013 (20130101); H04N
1/00037 (20130101); H04N 1/00053 (20130101); H04N
1/00063 (20130101); H04N 1/00087 (20130101); H04N
1/02815 (20130101); H04N 1/0284 (20130101); H04N
1/02885 (20130101); H04N 1/40056 (20130101); H04N
1/40068 (20130101); H04N 1/401 (20130101); H04N
1/121 (20130101); H04N 1/191 (20130101); H04N
2201/0412 (20130101) |
Current International
Class: |
H04N
1/028 (20060101); H04N 1/401 (20060101); H04N
1/00 (20060101); H04N 1/40 (20060101); H04N
1/12 (20060101); H04N 1/191 (20060101); H04N
001/04 () |
Field of
Search: |
;358/475,509,496,483,474,487,506 ;250/205,208.1,234,578.1,239
;362/800 |
References Cited
[Referenced By]
U.S. Patent Documents
Primary Examiner: Lee; Cheukfan
Attorney, Agent or Firm: Cesari and McKenna, LLP
Claims
What is claimed is:
1. An illuminator for a digitizing scanner a scanning camera
interconnected with a central processing unit, the scanning camera
defining discrete pixels arranged in a widthwise direction, the
pixels being adapted to acquire an image of a transparent film
passing through a widthwise field of view of the scanning camera,
the illuminator comprising: a plurality of light sources, each
being a light emitting diode and each being individually arranged
in a line across the widthwise field of view so as to face a first
side of the transparent film opposite to a second side of the
transparent film arranged to face the scanning camera; a plurality
of controllers each interconnected with respective of the plurality
of individual light sources, each of the controllers providing a
variable driving power to a respective of the plurality of
individual light sources; and wherein the central processing unit
is constructed and arranged to instruct the controllers to
respectively change the driving power to each of the plurality of
the light sources in response to a reading of intensity of
predetermined of the pixels that are mapped to each of the
plurality of light sources so that the scanning camera receives a
predetermined illumination pattern across the widthwise field of
view and wherein the controllers are adapted to respectively change
the driving power during a calibration procedure that occurs with
respect to operation of the digitizing scanner after startup of
operation of the scanner.
2. The illuminator as set forth in claim 1 further comprising a
housing for enclosing the light sources and a diffuser window
disposed between the light sources and the scanning camera.
3. The illuminator as set forth in claim 2 further comprising a
pair of walls approximately parallel to each other in the widthwise
direction.
4. The illuminator as set forth in claim 1 wherein the controllers
provide a plurality of current regulating circuits.
5. The illuminator as set forth in claim 4 further comprising an
intensity controller interconnected with each of the plurality of
controllers for varying a primary driving power to each of the
plurality of light sources, the intensity controller being
responsive to the central processing unit based upon a
predetermined maximum level of light exposure by each of the
plurality of light sources at the scanning camera.
6. The illuminator as set forth in claim 5 further comprising a
photodetector for detecting an output of a predetermined light
source of the plurality of light sources, the photodetector being
constructed and arranged to produce a reference signal, the
photodetector being interconnected with the intensity controller
whereby the predetermined maximum level of light exposure is set
and maintained based upon the reference signal.
7. An illuminator for a digitizing scanner having a widthwise field
of view, a scanning camera interconnected with a central processing
unit comprising: a plurality of light sources each individually
arranged in a line across the widthwise field of view; a plurality
of controllers, each interconnected with respective of the
plurality of individual light sources, each of the controllers
providing a variable driving power to a respective of the plurality
of individual light sources; wherein the central processing unit is
constructed and arranged to instruct the controllers to
respectively change the driving power to each of the plurality of
the light sources so that the scanning camera receives a
predetermined illumination pattern across the widthwise field of
view; wherein the light sources comprise a plurality of light
emitting diodes and wherein the controllers provide a plurality of
current regulating circuits; an intensity controller interconnected
with each of the plurality of controllers for varying a primary
driving power to each of the plurality of light sources, the
intensity controller being responsive to the central processing
unit based upon a predetermined maximum level of light exposure by
each of the plurality of light sources at the scanning camera; a
photodetector for detecting an output of a predetermined light
source of the plurality of light sources, the photodetector being
constructed and arranged to produce a reference signal, the
photodetector being interconnected with the intensity controller
whereby the predetermined maximum level of light exposure is set
and maintained based upon the reference signal; and wherein each of
the plurality of controllers comprises a digital/analog converter
and a common emitter buffer amplifier interconnected with one of
the plurality of light sources, the analog/digital converter being
interconnected with and receiving a digital control signal from the
central processing unit and interconnected with and receiving the
reference signal, the digital/analog converter having an output
interconnected with and controlling the common emitter buffer
amplifier whereby the driving power of the one of the plurality of
light sources is varied.
8. A method for calibrating an illumination assembly in a
digitizing scanner having a camera defining a widthwise line of
photosensitive pixels, the camera being adapted to scan widthwise
lines of an image in a lengthwise direction and a controller for
controlling the illumination assembly and the camera, the method
comprising the steps of: (a) during a calibration procedure that
occurs with respect to operation of the digitizing scanner after
startup of operation of the scanner, incrementally activating
predetermined light sources of a plurality of light sources, each
of the plurality of light sources comprising a discrete light
emitting diode, being arranged in a widthwise line to each project
an illumination light therefrom so as to transmit the light to a
side of a film being scanned toward the camera which faces an
opposing side of the film, the step of incrementally activating
including activating light sources that are each spaced apart by a
predetermined number of inactivated light sources until all light
sources have been activated; (b) identifying pixels on the camera
activated by each of the plurality of light sources and mapping
each of the activated pixels to a predetermined of the plurality of
light sources; (c) activating all of the plurality of light sources
at a predetermined power level and deriving an intensity signal
based upon projected light from the light sources at the camera;
(d) comparing an intensity value for each of the pixels identified
in the step of identifying with a predetermined reference intensity
value; and (e) changing a power level of each of the plurality of
light sources so that a measured exposure value of respective
pixels is closer to a desired exposure value based upon readings of
intensity at identified pixels mapped to each of the predetermined
of the plurality of light sources.
9. The method as set forth in claim 8 further comprising repeating
each of steps (c), (d) and (e) a predetermined number of times,
whereby the measured intensity values of each of the respective
pixels is made closer to a desired exposure characteristic.
10. The method as set forth in claim 9 wherein the step of
activating all of the plurality of light sources includes
controlling a driving power input to all of the light sources so
that each of the light sources has an intensity less than a
predetermined maximum intensity.
11. The method as set forth in claim 8 further comprising measuring
an output intensity of one of the plurality of light sources to
generate the reference intensity value.
12. A method for controlling a width of an illumination line
oriented in a widthwise direction and generated by an illumination
assembly in a digitizing scanner having a camera assembly for
receiving light from a scanned sheet that moves relative to the
camera assembly in a lengthwise direction transverse to the
widthwise direction, the method comprising the steps of: providing
a plurality of light sources oriented in a line along the widthwise
direction each of the light sources being directed to provide
illumination light that is centered at a predetermined position
along a widthwise line of the sheet; individually addressing
selected of the plurality of light sources to enable the selected
of the plurality of light sources to be deactivated; and
controlling the step of individually addressing so that the
selected of the plurality of light sources are light sources that
provide illumination light that is centered with respect to
portions of the widthwise line that are remote from a region of the
sheet desired to be scanned.
13. The method as set forth in claim 12 wherein the step of
controlling includes identifying widthwise edges of the sheet and
defining the portions of the widthwise line that are remote based
upon widthwise positions of the widthwise edges.
14. The method as set forth in claim 13 wherein the step of
identifying includes scanning the sheet to determine widthwise
locations of change in scanned density between a free space density
and a density greater than free space and the step of defining
includes deriving the widthwise positions of the widthwise edges
based the widthwise locations of the change in scanned density.
15. The method as set forth in claim 13 wherein the step of
controlling includes identifying widthwise limits of a region of
interest of an image on the sheet and defining the portions of the
widthwise line that are remote based upon widthwise positions of
the widthwise limits.
16. An illuminator for a digitizing scanner having a widthwise
field of view, a scanning camera interconnected with a central
processing unit comprising: a plurality of light sources, each
individually arranged in a line across the widthwise field of view;
a plurality of controllers, each interconnected with respective of
the plurality of individual light sources, each of the controllers
providing a variable driving power to a respective of the plurality
of individual light sources; wherein the central processing unit is
constructed and arranged to instruct the controllers to
respectively change the driving power to each of the plurality of
the light sources so that the scanning camera receives a
predetermined illumination pattern across the widthwise field of
view; a housing for enclosing the light sources and a diffuser
window disposed between the light sources and the scanning camera;
and wherein the wall comprise tapered walls, the tapered walls
tapering between a first spacing apart adjacent the diffuser window
and a second wider spacing adjacent the light sources, the walls
including a reflective surface thereon.
Description
FIELD OF THE INVENTION
This invention relates to an improved digitizing scanner, and more
particularly to a scanner for reading and storing graphical and
textual image data from transparent and translucent sheets such as
developed X-ray film.
BACKGROUND OF THE INVENTION
Electro-optical digitizing scanners are commonly employed as
peripheral devices linked with microcomputers and other data
processing and storage devices. Scanners enable graphical and text
data to be accurately converted into stored digital data for
further processing and interpretation by, for example, a
microcomputer. Scanners are adapted to read data from a variety of
media and formats. Opaque and transparent sheets are two common
forms of scanned media.
An image on a sheet is defined by light areas ("highlights") and
dark areas ("shadows"). To convert the light and dark areas into
corresponding image data, the scanner typically illuminates the
sheet with a light source. In one form of scanner, a camera
assembly moves along the length of the sheet. In another, the sheet
moves relative to a stationary camera. As the sheet moves relative
to the camera, the camera "scans" the width of the illuminated
image, converting the scanned portion of the image into a data
signal. This scanned image is said to be "digitized" in that the
image is converted into a data file stored in a digital format with
information representative of discrete segments or "pixels." The
data in the file includes instructions on how to assemble the
individual pixels into a cohesive two-dimensional image that
reflects the original scanned image. The data file also includes
information on the intensity value for each pixel and its color, if
applicable, or grayscale shade.
A common form of camera assembly for use in a digitizing scanner is
the solid-state CCD camera, which contains a linear array of
photosensitive picture elements, often termed "pixels." Each pixel
element receives light in its local area. The pixel generates an
intensity-based signal depending upon how much light it receives.
The aggregate signal of all the pixel elements is a representation
of a widthwise "line" of the image.
Generally, the CCD pixel array only scans a single line that is
several thousand pixels wide in the fast scan direction but that
has a height of only one pixel in the slow scan direction. The
array is typically wide enough to scan the entire image width at
once. Because an entire line is generally viewed at once, this is
known as the "fast scan" direction; since the delay is only in
downloading the signal from the CCD to the data processor.
Conversely, the direction of movement of the camera/image is known
as the "slow-scan" direction. In summary, images are scanned in a
"line-by-line" manner in which the image moves in the slow scan
direction relative the camera's fast scan field of view. As the
image passes through the field of view, a succession of scanned
width-lines of the image are converted into image data, and the CCD
element generates a continuous signal representative of the
intensity of each pixel in the line.
Scanners used for scanning opaque sheets must illuminate the image
by reflecting illumination light off the surface of the sheet from
the same side as the camera. Conversely, when transparent or
translucent sheets are scanned, the image is illuminated from the
opposite side of the sheet from the camera, allowing the light to
pass through the image to the camera. In this manner, the image
attenuates the light as it is transmitted through the sheet to the
camera.
CCD elements are generally smaller in width than the scanner's
total scan width. A focusing lens is employed to focus illumination
light from the scanned image onto the narrower viewing area of the
CCD. The focused image will generally exhibit a degradation in the
field of view at the far edges of the width (e.g. a loss of
exposure). This loss of exposure occurs because the amount of light
entering a lens tends to decrease at the edges of the field of view
according to the Cos.sup.4 characteristic of lenses. It is often
desirable to increase the light near the edges of the camera's
field of view to compensate for this effect. However, most
illuminators comprise only one or two discrete light sources, such
as a long fluorescent bulb. The intensity of such a bulb is not
generally controllable along its length. In fact the bulb may
exhibit variability in light output along its length, presenting a
different level of intensity to different pixels in the array. This
problem becomes exacerbated as the bulb ages. In addition, the
pixels of the CCD camera may exhibit different responses to the
same intensity of light. The CCD pixels can be calibrated to
account for most variations, but it is desirable to have the
capability of changing the profile of light presented to the
various pixels. In general compensation for an uneven light profile
is difficult using a single illumination bulb.
Scanners derive a large quantity of information from a single sheet
containing an image. When a sizable number of images are stored for
long-term use, superfluous data related to edges and margins can
become a concern. Substantial computing resources in both time and
storage capacity can be devoted to unneeded data. In particular,
images substantially narrower than the maximum field of view of the
scanner are often scanned as if the full width (in fast scan
direction) of the scanner is employed. It is desirable, therefore,
to accurately gauge the size of the needed data range, and only
scan the image within the needed range in both the fast scan and
slow scan directions. In the past this has been accomplished
primarily by manually inputting the size of the sheet to be
scanned. Alternatively, movable edge guides can be linked to a size
sensor that inputs the relative width of the input sheets. An
electromechanical/optical length sensor starts and ends the
scanning process as the front and rear edges of the sheet pass
through the scanner. However, these techniques still require
accurate registration of input sheets and do not determine the size
of the margins.
The scanning of translucent sheets is desirable in the medical
field, and presents particular challenges. In particular, there is
a need to digitally store and reproduce diagnostic radiological
films, commonly termed "X-rays." Most patient X-ray films, in fact,
are produced in a "series" that can consist of six or more
individual, interrelated X-rays. Hundreds, or even thousands, of
X-ray films are produced daily by a large hospital. By
electronically storing and indexing radiological images, they can
be made available indefinitely without taking up valuable physical
storage space. In addition, various specialized graphical processes
and image enhancement techniques can be used in connection with
stored X-ray images. Furthermore, scanned radiological data can be
easily transmitted to practitioners at remote locations via
electronic mail or facsimile. In all, the ability to accurately and
reliably scan developed X-ray film images provides an important
diagnostic tool for medical practitioners.
The scanning of developed X-ray film presents some particular
challenges. X-rays tend to exhibit a large area of shadows with
both abrupt transitions, and more subtle dark, clouded areas.
Hence, the CCD element intermittently must operate at a low output
level throughout the scanning process. Low light intensity causes
the CCD element to transmit a corresponding low output signal.
Electronic noise is accentuated at this low output level, causing
inaccuracies in the scanned image data. Incandescent and
fluorescent light sources often have short life spans that may
render them unsuitable for a large volume radiological scanner.
Alternatively solid-state illumination devices, such as light
emitting diodes (LEDs) must be used in large arrays. While they are
energy-efficient, long-lived, and consistent over their service
life, they may have wide variability in output intensity--even LEDs
in the same production batch. Thus the light intensity pattern
presented by an unadjusted array of LEDs can exhibit substantial,
undesirable variation in intensity across the scan width.
An illuminator that is larger in width that the image can cause
imaging problems. Scattered, stray light from the outer edges of
the illuminator, beyond the scanned width of the image, can cause
distortion and refraction patterns in the optics of the camera
assembly that degrade the scanned image. The width of projection of
most light sources, such as elongated fluorescent bulbs, cannot be
easily or reliable varied.
Medical X-ray film images are usually scanned at approximately 150
pixels per inch (PPI) resolution, since this value enables a
14-inch image to be displayed on a standard 2,000-pixel-wide
monitor. This resolution is generally considered sufficient for
radiological data storage and reproduction purposes. The native
resolution of many currently available CCD camera elements is
approximately 8,000-12,000 CCD picture elements. Divided over a
14-inch image this number of picture elements can provide native
resolutions of at least four times the number of pixels called for.
It is desirable to derive image data at the lowest needed
resolution to reduce scanning time and storage requirements. Lower
resolution is also desirable when transmitting data over low-speed
transmission lines to save time. An efficient technique for
changing the resolution of the system is desired.
Some circumstances may warrant the inclusion of specific image data
details at a higher resolution. These details are regions of
particular interest on an overall "parent" image. It is desirable
to provide a technique for producing higher resolution image files
of regions of interest, and electronically associating these
high-resolution detail files with the overall "parent" image.
While the individual pixels of currently available CCD camera
elements exhibit relatively consistent pixel-to-pixel output, there
is still signal variability between individual pixels in an array.
In particular, the signal for a dark image (the "dark current"
signal) can vary significantly from pixel to pixel. An adjustment
function is used to vary each pixel's output signal so that it
attains a desired uniform value. In particular, look-up tables
based upon predefined smoothing functions are often used to provide
a positive or negative adjustment bias to individual CCD pixels.
The gain exhibited by each individual CCD pixel is also adjusted by
deriving the change in output signal for a standard dark image and
a standard light image. The output signals of each of the thousands
of pixels in the array must be adjusted with an individual set of
bias and gain adjustment factors. This adjustment procedure
consumes substantial amounts of processing time and resources as
the linear output of each pixel is summed with an appropriate
positive or negative bias factor to provide each pixel with an
approximately equal dark current response. It is often desirable to
deliver a final signal from a CCD pixel in a logarithmic, or
another non-linear converted form. If the final non-linear signal
is to be sampled to provide the basis for adjusting the bias at the
linear input stage, then the device must have an accurate
representation of the function being used to convert the linear
signal into the final, non-linear signal. With foreknowledge of the
type of non-linear output to be expected for a given linear input,
the function can determine the mount of bias needed at the linear
stage to generate a proper shift in the output at the non-linear
stage. In other words, if the non-linear output is off by x, then
the function knows that a deviation of Log(x) has occurred at the
linear stage and this value is a correction factor to the
input.
However, preprogrammed tables of bias correction factors do not
always accurately predict the real response of a system. Likewise
many signal processing functions cannot be easily characterized.
Using a function or a preprogrammed table of expected correction
factors to effect CCD pixel bias calibration can result in
inaccurate data and can expend substantial time and computing
resources. This inaccuracy can be encountered even in linear signal
conversion circuits.
Notwithstanding the foregoing, reliability, repeatability of
results and ease of use remain a significant concern in any medical
device. In particular, a medical digitizing scanner must meet
certain guidelines promulgated by the US Food and Drug
Administration and other regulatory agencies in the United States
and abroad.
In view of the foregoing disadvantages of the prior art, it is an
object of this invention to provide a digitizing scanner,
particularly applicable to translucent sheets such as X-ray film
having an improved illumination system and camera arrangement that
produces highly accurate and consistent scanned image data. The
illumination system should be self-calibrating, have a long service
life and should compensate for optical and light source
inconsistencies. The camera arrangement should operate efficiently
at a desired resolution, should minimize distortion, exhibit a high
degree of optical precision and should include adequate noise
suppression capabilities for enhancing the quality of scanned
images. The camera element should be readily calibrated,
particularly in the logarithmic output signal domain. In addition,
the size of the image to be scanned should be accurately determined
and located automatically. The illuminator width should be readily
adjustable to fit the size of the image.
SUMMARY OF THE INVENTION
The digitizing scanner according to this invention overcomes the
disadvantages of the prior art by providing a plurality of improved
components and functions. In a preferred embodiment, the digitizing
scanner is generally arranged so that an illuminator transmits
light through a transparent or translucent sheet into a stationary
camera assembly. The image on each sheet passes through the field
of view of a linear CCD camera assembly as sheets are driven
lengthwise in the "slow scan" direction by a feed roller assembly.
The CCD captures a succession of lines of the image. Each line is
oriented widthwise, in the "fast scan" direction. Images are
transmitted as an image signal to the scanner's central processing
unit (CPU) and to a microcomputer or other data processing/storage
device as image file data. The sheets can comprise developed X-ray
film having black and white radiological images thereon. The CCD
can transmit information according to a corresponding black and
white "grayscale."
According to one embodiment, the scanner includes an improved
illuminator for illuminating an image. The illuminator consists of
a linear array of individually controllable light emitting diodes
(LEDs). The driving current for each LED is varied during a
calibration procedure in which the output light intensity of each
LED is independently measured by the scanner's camera, and the
driving current is adjusted in a series of adjustment cycles, or
"passes" to provide a predetermined consistent light output across
the array. The light output pattern naturally adjusts for inherent
optical and camera inconsistencies, since the output is varied
based upon the pattern actually viewed by the camera. The LED array
can include a photosensitive sensor that measures the light output
of one LED to derive a reference light intensity. The other LEDs in
the array are calibrated based upon this reference. A coarse
intensity adjustment can also be employed before each LED is
individually adjusted. Adjustment typically occurs in increments,
varying the LEDs driving current as a product of the old current
times the ratio of the average array illumination level versus the
LED's illumination level.
In another embodiment, a housing for the LED array can include a
pair of tapered walls that enclose part of each LED's bulb. The
walls taper to a narrower opening adjacent a translucent diffuser
window. The illumination light projected by the LEDs exits the
diffuser window in a highly diffuse form.
In another embodiment, a secondary illuminator is provided adjacent
the same face of the sheet as the camera assembly. The illuminator
can comprise a variety of acceptable light sources arranged to
project a reflected light onto a predetermined section of the
sheet, typically in a margin. The predetermined section includes an
opaque bar code strip or another identifier. The CPU can include
instructions for reading and interpreting the strip or identifier,
and can control the procedure for reading the strip at
predetermined times.
In another embodiment, the camera assembly comprises an enclosure
having a sealing window oriented in the widthwise direction for
receiving light transmitted from the image. The window allows light
to strike a series of reflectors that define an optical path. The
optical path terminates at a focusing lens and the CCD camera
element. A transparent covering window is positioned between the
focusing lens and the CCD camera element, adjacent the CCD camera
element. The sealing window is oriented at a non-perpendicular
angle to a plane passing perpendicularly through the optical path
to divert stray light out of the optical path. In one embodiment
the angle is set preferably at 7.degree.-15.degree.. However, any
angle that enables diversion of stray light without unduly
compromising the optical performance of the camera assembly is
acceptable. The camera and the covering window, as a unit are
tilted in the housing at an angle preferably between 7.degree. and
15.degree. relative to a plane passing perpendicularly through the
optical path. The reflectors can be mounted on a rigid frame member
on respective adjustable mounts.
In another embodiment, the output image signal from the CCD element
can be processed dynamically to reduce noise in the low-intensity
(shadow) signal range. A two stage logarithmic amplifier is
employed to amplify the signal by 100 dB in two 50 dB stages. A
variable low-pass filter reduces the bandwidth of the signal
between the two stages according to predetermined criteria.
Specifically, a control amplifier controls the filter's maximum
allowable signal bandwidth based upon the current value of the
output image signal. For low-intensity output signal values below a
predetermined lower limit, a minimum allowable bandwidth is
selected. The allowable bandwidth increases to a maximum value
wherein a predetermined upper limit is reached. This upper limit is
at the upper end of the low-intensity output signal value range.
The filtered output signal of the filter is passed through the
second stage of the logarithmic amplifier and the output of the
second stage amplifier is summed with the output of the first stage
logarithmic amplifier to produce a 100 dB filtered logarithmic
output signal. This signal is converted into useable digital and
linear form by appropriate converters.
In another embodiment, resolution of the CCD camera element is
reduced from a higher resolution by averaging the values of
adjacent fast scan pixels in the fast scan direction and deriving a
single pixel intensity for the entire grouping. Pixel intensity
values are preferably combined in adjacent pixel groupings of 2, 4
or 8. The resulting summed intensity values are averaged in the
binary domain by shifting the sum by 1, 2 or 3 bits, respectively.
Averaging of pixel values in the slow scan direction is
accomplished by varying the scan speed of the image to present a
plurality of lines to the CCD camera array in a given scan cycle.
In a preferred embodiment, the scan speed is varied by controlling
the operating speed of the feed roller drive motor. The CCD camera
element samples lines at a fixed rate. By increasing the scan
speed, a larger area in the slow scan direction is presented to the
camera during each sample cycle. The area scanned is read by the
CCD camera element as an average intensity signal for each CCD
pixel. Preferably, 2, 4 or 8 lines are averaged, resulting in an
average line signal that represents a line of pixel intensity
values for the entire grouping of lines. The average pixel values
derived from either, or both, fast scan and slow scan pixel
averaging are stored as an image data file according to the new,
reduced resolution.
In another embodiment, the size of a sheet fed into the scanner is
automatically determined, and the amount of data taken and stored
by the scanner is adjusted using the intensity readings of the
camera element. The sheet is fed by the rollers into the field of
view of the camera assembly. The camera begins scanning before the
sheet arrives at its field of view. The intensity transition
between the free space before the edge of the sheet and the
attenuated intensity as the sheet passes into the field of view is
identified by the CPU as the lead edge of the sheet. The CPU maps
the location of the lead edge to the feed motor's position by
counting steps or reading another movement sensor signal
operatively connected to the motor. The widthwise edges of the
sheet are then determined by locating intensity transitions on
either side of the sheet along the fast scan direction. The
location of the side edges can be mapped at predetermined intervals
relative to the motor's position or the location can be mapped
continuously. The CPU continuously polls for a second intensity
transition at the tail edge in the slow scan direction. The second
transition, when identified by the CPU, is mapped relative to the
motor's location by the CPU and the sheet is reversed by the
rollers until the top margin is again upstream of the camera
assembly's field of view. The sheet is then fed by the rollers
through the scanner again. Only data falling substantially within
the mapped boundaries of the sheet, based upon the current position
of the motor, are acquired and stored for further processing.
In another embodiment, the illumination assembly is adjustable to
deactivate selected light sources having respective centers of
projection that fall outside of the widthwise edges of the sheet,
or another set of widthwise limits. The programmable current
sources for selected light sources are instructed by the CPU to
assume a minimum current or "off" setting. The deactivated light
sources can be selected based upon their known physical locations
along the width of a sheet. Selection can occur based upon manually
input width measurements or based upon automatic size sensing
functions, such as the procedures described above.
In another embodiment, the image is annotated with stored
identifiers that link the image to a higher resolution file having
image data related to a specific area of interest on the main
image. The linking can be made according to the ANSI-DICOM-3
standard. According to this standard the main image is stored in a
predetermined format. A high-resolution subfile is created by
rescanning a particular region of the sheet containing the main
image. This subfile is also stored in the predetermined format, and
the two files are linked for subsequent display and data
transmission. The use of a smaller high-resolution linked file
saves valuable storage capacity and data handling time,
particularly during data transfer to remote sites. Sheets remain in
the feed rollers of the scanner until all scans have been
accomplished, selectively driving the sheet in reverse, and forward
again until all scanning operations have been completed.
In another embodiment, a method for bias calibration of CCD pixels
in the camera array that enables efficient adjustment of the bias
of individual CCD pixels in the linear mode based upon a sensed
output of each of the pixels in the logarithmic domain is provided.
A group of pixels from the overall array is exposed to a
substantial absence of light representing a maximum dark intensity
image. This is accomplished by deactivating the illumination
assembly. The linear output of each of the pixels is amplified
using a signal converter, such as a logarithmic amplifier, and more
particularly, the two-stage amplifier employed in the
above-described filtering circuit. Each CCD pixel in the group is
assigned a specific bias value that is summed with the respective
pixel's output to the dark intensity to create an incremental ramp
of bias-adjusted linear inputs to the logarithmic amplifier that
are amplified into a set of logarithmic system response values at
the output of the logarithmic amplifier. The ramp of individual
bias values is stepped incrementally from a minimum negative bias
value to a maximum positive bias value. The minimum negative value
and the maximum positive bias value are of equal magnitude and
opposite sign of voltage/current according to a preferred
embodiment. The bias values between the minimum and maximum are
equal, increasing increments of voltage/current. The logarithmic
amplifier is precalibrated to generate a minimum, negative system
response output in the logarithmic domain when the minimum bias
value is input and a maximum, positive system response output in
the logarithmic domain when the maximum bias value is input. The
negative and positive system response output values are, likewise,
equal in magnitude, and opposite in sign according to a preferred
embodiment. Each logarithmic domain system response is mapped to
the input bias that produced the response. The mapping process
results in a table or "curve" of logarithmic system response versus
input linear bias. The approximate middle of the curve represents a
desired system response of 0 for the CCD array. A given bias value
on the ramp (approximately half-way between the minimum and maximum
value) produces a baseline, 0-system response. The values for data
points in the curve can be represented by numerical digital
integers that are translated using appropriate digital/analog and
analog/digital converters. The curve of system response versus bias
is manipulated through inversion and translation to derive another
curve (in look-up table form) of bias adjustment factors for a
respective set of deviation values from the desired base response
for a pixel at a dark current output. The response of each pixel in
the array is then measured in the logarithmic domain. The curve is
queried to assign the appropriate linear bias adjustment factor to
each pixel based upon its logarithmic output. The bias adjustment
factor assigned to each pixel is mapped to that pixel and applied
to the linear output of that pixel each time it transmits an
intensity signal. According to a preferred embodiment, eight bias
and associated system response data points can be summed to
generate a single look-up table point. Deviations from the base
value that fall between the averaged points can be derived through
linear interpolation. By assigning a ramped bias to each pixel in a
group that spans approximately one decade of logarithmic output, a
bias correction factor table can be developed in a single
8-millisecond scan cycle. Curve-smoothing and point-averaging
functions are employed to ensure that variations in the output
signals of individual pixels do not unacceptably disrupt the
continuity of the response curve.
It is expressly contemplated that any of the above-described
embodiments can be employed in conjunction with one or more of the
other above-described embodiments in the digitizing scanner
according to this invention.
BRIEF DESCRIPTION OF THE DRAWINGS
The foregoing and other objects and advantages of the invention
will become more clear with reference to the following detailed
description as illustrated by the drawings in which:
FIG. 1 is a schematic cross-section and perspective view of a
digitizing scanner for scanning and storing image data from
transparent and translucent sheets according to this invention;
FIG. 2 is a plan view of an exemplary translucent sheet comprising
a developed X-ray film;
FIG. 3 is a schematic broken side view of an LED array illumination
bar for use in the scanner of FIG. 1;
FIG. 4 is a side cross-section of a reflector and diffuser assembly
for the LED array illumination bar of FIG. 3;
FIG. 5 is a block diagram of a control module for the LED array of
FIG. 3;
FIG. 6 is a schematic circuit diagram of an intensity control
amplifier for each LED in the array of FIG. 3;
FIG. 7 is a schematic circuit diagram of a programmable current
source for each LED in the array of FIG. 3;
FIG. 8 is a flow diagram depicting an LED array calibration
procedure according to this invention;
FIGS. 9, 10 and 11 are schematic broken plan views of the LED array
illumination bar of FIG. 3 detailing an incremental shift procedure
for calibrating the intensity of each LED;
FIG. 12 is a more-detailed flow diagram of an LED-intensity
adjustment procedure for the calibration procedure of FIG. 8;
FIGS. 13-17 are graphs that showing, progressively, the output of
each LED in the bar array at various stages of the calibration
process;
FIG. 18 is a somewhat schematic side cross-section of the camera
assembly for the use in the scanner of FIG. 1 detailing the
reflector and lens arrangement;
FIG. 19 is a graph showing the presence of electronic noise at
low-light response levels in an unfiltered CCD camera circuit
according to the prior art;
FIG. 20 is a block diagram of a dynamic noise suppression circuit
for use with a CCD camera circuit according to an embodiment of
this invention;
FIG. 21 is a graph showing reduced electronic noise at low-light
response levels in the CCD camera circuit having noise reduction
according to an embodiment of this invention;
FIG. 22 is a schematic plan view of an array of pixels grouped for
intensity-averaging according to an embodiment of this
invention;
FIG. 23 is a flow diagram of a CCD pixel intensity-averaging
procedure in the fast scan direction according to an embodiment of
this invention;
FIG. 24 is a flow diagram of a CCD pixel intensity-averaging
procedure in the slow scan direction according to an embodiment of
this invention;
FIG. 25 is a flow diagram of an automatic image size sensing
procedure according to an embodiment of this invention;
FIG. 26 is a flow diagram of an illumination assembly width control
procedure according to an embodiment of this invention;
FIG. 27 is an exemplary developed low-resolution X-ray film image
referencing an associated high-resolution detailed image according
to an embodiment of this invention;
FIG. 28 is a flow diagram of a procedure for generating a
high-resolution detailed image and associating the image with a
portion of a low-resolution overall image as shown in FIG. 27;
FIG. 29 is a block diagram of the output stage of a pixel in the
CCD array according to an embodiment of this invention, detailing
the preamplification, amplification, analog/digital conversion and
bias adjustment stages;
FIG. 30 is a flow diagram of a procedure for calibrating bias of
each pixel in the CCD array having an output stage arranged
according to FIG. 29; and
FIGS. 31-33 are graphs of curves derived according to the procedure
of FIG. 30 for providing bias adjustment factors to each pixel of
the CCD array.
DETAILED DESCRIPTION 1. System Overview
The scanner of this invention is shown in overall detail in FIG. 1.
The scanner system 30 comprises an outer housing 31 constructed
from metal, plastic or another acceptable, housing material that
covers. An internal framework (not shown) is employed to support
the housing and the components in a predetermined alignment. Within
the housing is mounted a feed tray 32 holding a plurality of sheets
in a stack 34. The tray 32 moves upwardly and downwardly (double
arrow 36) to feed sheets intermittently to two pairs of feed
rollers 38 and 40. The feed roller pairs each comprise a respective
drive roller 39 and 41 and a pressure roll 49 and 51. A stripper 42
and a pick roll assembly 44, in combination, singulate each
uppermost sheet in the stack 34, and direct each sheet into the
feed roller pairs 38 and 40. The operation of the feed mechanism
can be conventional or, alternatively, single sheets can be fed
directly from a feed slot to the feed roller pairs 38 and 40. The
lower drive rollers 39 and 41, respectively of each feed roller
pair 38 and 40 are driven by a central drive motor 46 and each
individual drive roller is interconnected with, and driven by, the
drive motor 46 by a series of belts 47 that can be conventional.
The motor 46 can comprise a stepper motor that drives the rollers
39 and 41 in either a forward or reverse direction based upon a
predetermined number of "steps" or rotational increments.
An upstream sensor 48 resides at the inlet of the upstream feed
roller pair 38 and another, downstream sensor 50 resides at the
outlet of the downstream feed roller pair 40. The upstream sensor
48 is also termed the "trail edge" sensor since it is used herein
to sense passage of the trailing edge of a sheet therethrough.
Likewise, the downstream sensor 50 is termed the "lead edge" sensor
since it is used herein to sense the arrival of the lead edge to
the sensor's position in the feed path. Each edge sensor 48 and 50
can comprise a conventional microswitch or electro-optical
transmitter having an output connected to the scanner's Central
Processing Unit (CPU) 52.
The CPU 52 can comprise a variety of data processing and control
including one or more microprocessors. The function of the CPU is
described further below with reference to each individual component
and its functions. Each operating component is, likewise,
interconnected with the CPU by an appropriate analog,
digital-serial or digital-parallel link. Power to the CPU and other
components is provided by a power supply 54 that receives power
from batteries, an alternating current (AC) source or another
acceptable electrical current source. The central drive motor 46 is
linked with the CPU 52, and receives speed, direction and on/off
commands from the CPU. An exemplary sheet 60, which comprises a
translucent developed X-ray film, is driven by the feed rollers 38
and 40, under control of the CPU 52 by the motor 46. The motor 46
is capable of both forward driving motion (arrow 61) and, opposing,
rearward driving motion upon command. The pick roll assembly 44 is
driven by a separate CPU-controlled motor (not shown).
FIG. 2 illustrates generally the exemplary developed X-ray film
sheet 60. A primary radiological image 62 is shown in the center of
the sheet. The image 62 is largely transparent with various dark,
opaque, inclusions 64 and 66 within its boundaries. In general, the
surrounding background 68 is dark and mainly opaque. The background
68 typically extends outwardly to the perimeter edge 74 of the
sheet 60.
With reference again to FIG. 1 the sheet 60 is illuminated by an
illumination assembly 76. Illumination assembly 76, as described
further below, comprises an array of semiconductor light emitting
diodes (LEDs) 78 that emit a nearly monochromatic light. The light
is directed by a reflector assembly 80 and through a diffuser
window 82. It passes through the sheet 60 as a diffuse beam 84
where it is received by the camera assembly 86. While not shown, a
transparent platen can be located between the sheet 60 and the
camera assembly 86. The camera assembly 86 of this embodiment is
enclosed in a rigid sheet metal enclosure 88. An elongated (taken
in a direction perpendicular to the page of FIG. 1) window 90
enables the image light 84 passing through the sheet 60 to enter
the camera assembly enclosure 88. The light is reflected along an
optical path through a series of reflectors 92, 94, 96 and 98, and
through a focusing lens 100 to a CCD camera element 101. The CCD
camera element 101 receives a line image that has been focused by
the lens 100 onto the row of picture elements (CCD "pixels")
extending across the element's width (taken in a direction
perpendicular to the page of FIG. 1). Specifically, the CCD camera
assembly includes a line of photosensitive pixels that each
individually respond to light intensity with a corresponding
electronic intensity signal. The CCD camera element 101 used
according to a preferred embodiment of this invention comprises a
Kodak KLI8013 grayscale CCD unit having approximately 8,000 CCD
pixels in a line. Each CCD pixel has a size of approximately 9
microns square. The lens 100 focuses the 14-inch wide line produced
by the illumination line onto a CCD element line. The lens 100 and
reflectors 92, 94, 96 and 98 define an optical path that results in
a viewed image pixel size of approximately 0.00171 inch square. In
other words, each 9-micron CCD pixel views a 0.00171 square piece
of the image, based upon the lens' ability to focus the image onto
the smaller CCD element. The term pixel will be used
interchangeably to describe image pixels having a 0.00171-inch size
and CCD pixels having a 9-micron size. The resulting inherent
resolution of a full size, 14-inch wide image is approximately 584
pixels per inch (PPI). An image is sampled by the line array every
8 milliseconds, requiring the drive to feed the sheet at
approximately 0.2 inch per second (152 steps of the drive motor 46
in this embodiment) to ensure that the pixels are presented with a
new line of the image each time an image is sampled. Data collected
by the pixel line array CCD camera element 101 is transmitted
through a data line 103 to the CPU 52 for processing.
Following a series of processing steps, the CPU transmits acquired
image data over a communication line 105 to a microcomputer 107.
The microcomputer 107 processes the image data according to known
digitizing procedures and stores the image data in an appropriate
data storage device such as a disk 109. Data can be manipulated
using a user interface that can include a keyboard 111 and a mouse
113 and can be displayed on a monitor 115. Stored image can also be
downloaded to other computers over a network or digitally reprinted
using a telephone-connected facsimile machine. II. Illumination
Assembly
A. Array Structure and Control Circuit
As discussed generally above, illumination is provided by a "bar"
assembly having between 58 and 60 individual LEDs spaced across a
14-15 inch width at approximately equal spacing of 0.25 inch. A
broken view of the LED bar array 100 having a plurality of
individually spaced LEDs 102 is depicted in FIG. 3. In this
embodiment, commercially available, red-emitting, high-output LEDs
are employed. Such LEDs are rated to operate at an input current of
approximately 50 mA with an approximate input voltage of 2.5 V. The
output intensity can be controlled through a wide range by varying
the input current. Each LED is individually powered by a wire pair
(not shown) mounted to a circuit board 106 that contains the
control circuitry described further below. The CPU 52 provides
control signals to the LED array circuit board 106 through a
control cable 104. As also described further below, a commercially
available, variable output phototransistor 108 is located adjacent
one of the centrally located LEDs (LED 32) in the array 100. It
transmits a signal that is proportional to the light output from
LED32.
The LED bar array 100 is mounted in an illuminator housing 110 as
detailed specifically in FIG. 4. The housing 110 can comprise a
sheet metal channel or another suitable enclosure. The precise
shape of the housing can be varied depending upon the geometry of
the scanner interior. The LEDs 102 are directed to transmit light
downwardly into a pair of elongated (perpendicular to the page of
FIG. 4), tapered walls 112 that run approximately parallel to each
other along their direction of elongation. In this embodiment, the
LED bulbs each project partially into the enclosure of the walls
112. The walls 112 define an opening 114 at their tops to receive
the line of LEDs 102, and the walls 112 are enclosed at their
bottom by a translucent diffuser window 116. The diffuser window
116 is constructed from a frosted glass or synthetic such as
plexiglass to project a highly diffuse light from the window 116
(see arrows 118). The inner surfaces 120 of the walls 112 are
provided with a reflective film that further directs light
projected from the LEDs 102. In this embodiment, the walls stand at
a height H1 of 1.375 inches from the diffuser window 116. Their
distance apart W1 at the opening 114 is approximately 0.25 inch and
their distance apart W2 adjacent the diffuser window 116 is
approximately 0.125 inch. The width and thickness of the window 116
can be varied depending upon the structural characteristics of the
housing 110, but it is, typically, at least as wide as the wall
spacing W2.
Referring to FIG. 5, the LED array is operated by a LED control
circuit 130. This circuit provides a regulator constant current
supply for each of the 58 individual LEDs of this embodiment. Each
LED is individually adjustable. It is not uncommon for LEDs to vary
in intensity as much as four times relative to one another.
Generally, the circuit comprises an overall intensity control
amplifier 140 that receives a lamp control signal LMP_CTL of
between -2.5 and +2.5 volts. The intensity control amplifier
regulates the overall input voltage to the entire array 100. The
intensity control amplifier 140 also receives an input signal from
the phototransistor 108 adjacent the 32nd LED in the array (LED32
in FIG. 3).
The intensity control amplifier transmits an overall voltage signal
to the programmable current source 150 which comprises a group of
corresponding individual digital/analog converters that generate a
controlling output voltage for each respective LED. The
programmable current source, using the intensity control amplifier
signal as a reference, modifies the driving voltage for each
individual LED in the array 100 based upon control input signals
received from the CPU through a digital decoder 160. The digital
decoder enters control data through a data line 162 and selects the
LED's specific digital/analog converters using a select/enable line
164. An LED32 current detector 170 generates a lamp monitor signal
LMP_MON that returns a measurement of intensity for LED32 to the
CPU 52.
As detailed in FIG. 6, the intensity control amplifier 140 resides
on the array circuit board 106. The intensity control amplifier 140
comprises a differential amplifier 180 that receives inputs from
the lamp intensity control signal LMP_CTL and also from a reference
signal LMP_RTN. LMP_RTN is generally a grounded signal, designed to
correct for transmission losses experienced by the control signal
LMP_CTL. The output of the differential amplifier 180, comprising a
corrected control signal is provided to an integrating difference
amplifier 182. The output of the phototransistor 108 is compared to
the resulting intensity control signal by the amplifier 182, and
the resulting value passes through a common emitter transistor
buffer 184 which is sufficient to provide an adequate current to
drive all of the digital/analog converters in the array. The output
signal 186 of this buffer 184 is routed to the various
digital/analog converters on the circuit board 106 of the LED
array.
The digital/analog converters are part of the programmable current
source block 150, which is further detailed in FIG. 7. The signal
186 from the intensity control amplifier is provided as a reference
voltage to each digital/analog converter 190. A three-bit data line
162 provides individual LED control signals from the decoder 160
(FIG. 5) and serial interface of the CPU 52. A specific enable
signal EN is also provided by the decoder through line 164 to
activate each digital/analog converter 190. Typically, eight
digital/analog converters are provided on a single integrated
circuit chip package. Seven digital/analog converter chips are used
in this embodiment, with the first two LEDs (LED 1 and LED2) and
the last two LEDs (LED57 and LED58) operating in tandem. This
arrangement, however, is chosen merely based upon cost
considerations, and it is contemplated that each and every LED can
be operated by a discrete digital/analog converter according to
this invention. Likewise, groups of LEDs can be operated by a
single digital/analog converter in an alternate embodiment.
The digital/analog converter 190 for each LED generates an output
voltage V.sub.out, based upon the reference voltage V.sub.ref that
is multiplied by an 8-bit digital word received from the data line
162. In other words V.sub.out =V.sub.ref (CONTROL/256) where
CONTROL is a digital value between 0 and 255. A pair of transistors
194 configured to form a common emitter buffer amplifier receive
the V.sub.out signal. Using a 49.9 ohm set-resistor, a variable LED
driving current I.sub.LED is provided to the LED 102. I.sub.LED
=V.sub.out /49.9 according to this relationship. Hence, a V.sub.out
equal to 0.625 volts produces a nominal LED current of 12.5 mA. A
12.5 mA current is the target value desired for initially
calibrating LED32.
As discussed above, it is not uncommon for LEDs to have a
brightness/intensity variation of as much as 4:1 for any two LEDs
in a grouping. This variation can result in wide deviations in
illumination characteristics for a given array of LEDs. See, for
example, the distribution curve 200 of FIG. 13 (described further
below) which shows relative intensity values for a given array of
unadjusted LEDs all driven at constant current and voltage in a
scanner according to this invention. The curve 200 shows wide
variability between adjacent LEDs. Note that the curve 200, and the
other calibration data derived herein are based upon the light
pattern received by the CCD camera assembly 86 (FIG. 1) used for
the scanner of this embodiment.
Note that, for a given intensity of light, the individual CCD
pixels may exhibit a non-uniform response. In other words, the
output signal of each CCD pixel may be slightly different due to
manufacturing variations in each CCD pixel. Some variations in
individual CCD pixel output are compensated for by calibrating the
CCD camera element output. The architecture of the LED array
enables further adjustment for nonuniformity in CCD pixel output
and in nonuniformities in the optical system by allowing the local
intensity of light in the scanned line to be raised and lowered.
The ability to adjust the local intensity of light, by adjusting
individual LEDs is used, for example, to compensate for the loss of
exposure at edges of the image width. As noted above, the focus
lens 100 generates the above-described COS.sup.4, wherein LEDs near
the outer edges of the field of view (e.g., LEDs near LED1 and
LED58 in FIG. 3) are viewed as dimmer by the CCD element. The
calibration procedure to be described below, in fact, compensates
for such inherent variability in the optical system by raising the
output of the LEDs near the outer edges. By adjusting the
individual LEDs to desired levels, the CCD pixels each transmit an
intensity signal that falls within a uniform range.
B. LED Array Calibration and Adjustment Procedure
Referring specifically to FIGS. 8 and 12, an illumination assembly
calibration procedure is detailed. Reference will also be made,
generally to FIGS. 13-17 which detail an actual measured response
curve for an illumination assembly of this embodiment as viewed by
the CCD camera assembly of this invention. Upon scanner system
start-up, the illuminator activates all LEDs in step 250. Start-up
can comprise a simultaneous ramping-up of the digital control input
to each of the LED digital/analog converter circuits in each
respective programmable current source until transition in each
digital/analog converter has occurred. Analyzing the output
received by the CCD camera assembly, the CPU inputs an intensity
control signal LMP_CTL to adjust all LEDs so that the highest value
CCD pixel intensity does not exceed a predetermined maximum value
in step 250. In this embodiment, a measured CCD pixel output value
of "4,000" (as recognized by the Kodak KLI8013 grayscale CCD unit)
is considered the maximum allowed intensity value. The value 4000
is considered to represent a pure "white" output on the CCD
grayscale (described further below). The maximum intensity value
can be varied based upon the make and type of camera element being
used. As this coarse adjustment stage of the calibration procedure,
all LEDs are operated at a substantially equal, coarsely adjusted
input current and voltage to their respective programmable current
sources.
Subsequent to the coarse adjustment step 251, the individual LEDs
in the array are located based upon where they are viewed within
the CCD element. In other words, each of the illuminated LEDs is
mapped to a particular group of pixels in the line array of the
CCD. Specifically, the mapped pixels are averaged to determine the
center line or "centroid" for each LED. The procedure for mapping
LEDs to pixels is described in steps 252 and 254 and in decision
block 256 of FIG. 8, which are described further below.
FIGS. 9, 10 and 11 schematically illustrate the operation of the
LED array bar during the centroid location procedure outlined in
main calibration procedure steps 252, 254 and 256. To determine
which pixels correspond to each LED center line, LEDs are switched
on and off such that only LEDs spaced eight positions apart from
each other are on at any given time. This eight-LED spacing is
generally sufficient to reduce "crosstalk" between light patterns.
In other words, discrete groups of pixels on the CCD element will
receive the light transmitted by each spaced-apart LED without
receiving substantial light from adjacent LEDs. In FIG. 9, LED4,
LED12 and a succession of further spaced-apart LEDs up to and
including LED52 are switched on. Each LED outputs light at the
coarsely adjusted maximum intensity level. The CPU reads the
location of the pixels showing the maximum intensity for this set
of LEDs. The CPU determines the pixels outputting the maximum
intensity and correlates these intensity readings with the
"location" of the particular LEDs. This location on the CCD array
is stored as the centroid position for the LED being addressed. The
CPU assumes that LEDs will illuminate pixels in the same spacing
pattern as the LEDs are positioned in the bar. Hence LEDs are
identified in an order corresponding to the order of illuminated
pixels in a widthwise row across the CCD array (e.g. leftmost
illuminated pixels must be LED4, next to the right must be LED12,
and so forth . . . .)
Next, in FIG. 10, the illuminated group shifts so that LEDS, LED13
and the succession of spaced-apart LEDs up to LED53 are switched
on. The CPU again records the centroids corresponding to the
addressed LEDs as described above.
In FIG. 11, LED6, LED 14 and a succession of spaced-apart LEDs up
to, and including, LED54 are then switched on, and the
corresponding pixels are identified. This process continues through
eight iterations until all LEDs have been correlated to particular
pixels in the CCD element. When the cycle is complete, the decision
block 256 (FIG. 8) routes the procedure to the next step in the
main illuminator calibration process.
In step 258 of the main calibration process (FIG. 8), all LEDs are
again illuminated, and the output of LED32 is read using the
phototransistor 108 (FIGS. 3 and 6). This output value is used to
establish a baseline emission value for the given current input to
LED32. LED32 is deemed to provide an appropriate level of light
output/intensity for the array, as a whole. The output current
generated by the phototransistor in response to LED32 is translated
by the current detector 170 (FIG. 5) into the signal LMP_MON. The
emission value of the phototransistor is used as a reference by the
intensity control amplifier 140 (FIGS. 5 and 6). Based upon this
reference, all LEDs are now readjusted in steps 260 and 262 and
decision blocks 264 and 266.
The adjustment of the output intensity of individual LEDs in step
262 occurs in a series of passes. Note that each LED is assigned an
8-bit output level according to this embodiment, enabling 256
different settings for each LED. Passes are employed since each LED
will have a certain amount of effect on the measured intensity of
adjacent LEDs. The effect of each LED on its neighbors varies
depending upon (1) the angle (however slight) at which each LED is
slanted; (2) the optical characteristics of each LED-how it
projects its light; and (3) inconsistencies in the bar's diffuser
window and tapered, reflective walls. In this embodiment, decision
block 264 is instructed to end the procedure after eight passes
which is considered sufficient to adequately smooth variations in
intensity output. Prior to eight passes decision block 266
determines whether further adjustment is required or if all LEDs
have a measured intensity that falls within a predetermined range.
For each pass, the adjustment of the individual digital/analog
converters for each LED in step 262 occurs according to the
adjustment procedure more-particularly detailed in FIG. 12. First,
the intensity value for each LED is computed by averaging the value
for its adjacent LEDs in step 280. The value for each LED, again,
can be determined since the CCD element pixel corresponding to the
centroid for each individual LED are known the intensity presented
to those pixels can be measured. Using the measured intensity value
for each LED, as derived from the averaging processing in step 280,
the average illumination of the entire array is computed in step
282. Finally, the current input to each LED can be varied (via its
respective digital/analog converter) according to the following
relationship set forth in step 284:
Referring to FIGS. 14, 15, 16 and 17, the respective curves 290,
292, 294 and 296 show the relative response of the CCD element to
the individual adjustment of each LED in the array through one,
two, three and four respective passes of the procedure. Clearly,
the measured response of the CCD element is flattened substantially
as the measured output of each LED is equalized with the desired
standard intensity value. Further passes (not shown) result in even
lower variation between LED intensities up to the maximum of eight
in this embodiment.
Once the calibration of LEDs relative to each other is completed in
step 258, the main calibration procedure (FIG. 8) proceeds to
decision block 300. The procedure queries whether multiple passes
have occurred. If only a single pass has occurred, all LEDs are
adjusted upwardly to their maximum value in block 302 and
calibration step 258 is performed again. The ensures that the array
has been properly calibrated. It is possible that only one pass was
taken because the outputs of the LEDs were too dim to obtain an
adequate measurement. If more than one pass has occurred, then the
decision block 300 branches to the white level adjustment step 304.
The intensity control amplifier 140 (FIGS. 5 and 6) is adjusted by
the CPU so that an appropriate white level is attained by the pixel
array as a whole after it has been adjusted. Again, a pixel
intensity value of 4,000 (corresponding to a 5-volt digital/analog
converter value as recognized for the commercially available Kodak
CCD unit of this embodiment) is desirable according to this
embodiment. Following the white level adjustment step 304, overall
system calibration is undertaken in step 306 in which the newly
adjusted illumination level is stored by the CPU as maximum white
and the darkest level is stored as black. Maximum black is
generated by shutting down the LED array. An advantage to using an
LED array is that it can be quickly shut down and reactivated. All
intensity values in between maximum light and maximum black are
read at the CCD output with 12-bit resolution.
It is contemplated that further detection devices and procedures
can be provided through the illumination assembly of this
invention. For example, calibration of individual LEDs can occur
one-at-a-time by cycling each LED on and off in succession and
measuring the relative intensity. In this manner, the effects of
adjacent LEDs are minimized and failures in individual LEDs can be
readily identified. However, calibration with all LEDs
simultaneously activated is preferred because cross-talk between
adjacent LEDs may be non-uniform. By activating all LEDs at once
the effect of non-uniform cross talk can be accounted for in the
adjustment process. Several iterations or "passes" can be required
to adjust the LED array according to this embodiment to address the
effect of this cross-talk by "smoothing" the overall light-output
curve of the array.
C. Optional Illuminator and Function
With reference again to FIGS. 1 and 2, an optional illuminator 310
is located between the sheet 60 and the CCD camera element. The
illuminator 310 is oriented at a arbitrary angle relative to the
perpendicular beam 84 entering the CCD element from the illuminator
assembly 76. This orientation prevents the illuminator 310 from
interfering with the beam 84. In particular, the illuminator 310
should be oriented to reflectively illuminate a predetermined width
of the side edge of the sheet 60. The sheet 60 includes, along the
right side of the upper edge 312 an opaque bar code strip 314. It
is often desirable to provide opaque, self-adhesive bar code strips
and other identifiers to developed X-ray film and other translucent
or transparent sheets for identification purposes. These strips are
typically added after the film is developed. The illuminator 310 is
directed to selectively illuminate this strip. When the sheet is
placed with the strip face down (e.g., facing the camera assembly
86) the array illuminator 76 cannot effectively illuminate the
opaque strip 314. Hence, the illuminator 310 provides a reflected
light from the same side as the strip 314. The CPU controls the
illuminator so that it provides the light as the front edge 312
passes into the field of view of the camera assembly 86. A
digitizer within the CPU or microcomputer can store and decode the
information contained on the strip 314 using known procedures for
analyzing graphical, text and bar code data. In the above-described
manner the illuminator light does not interfere with the reading of
the main body of the film, and the read information is located on a
marginal part of the film that is largely unused and that does not
obscure the central image. The illuminator can be constructed from
one or more LEDs or from another light source such as a neon or
halogen incandescent bulb. The projected light should be sufficient
so that an opaque strip can be read through its reflected light.
III. Camera Assembly
A. Camera Assembly Structure and Function
The camera assembly 86 (FIG. 1) is provided a self-contained unit
according to this embodiment. Referring to FIGS. 1 and 18, the beam
of image-attenuated illumination light 84, which generally defines
a thin, elongated projection (elongated in a direction
perpendicular to the page and corresponding to the width of the
illuminator) first passes through an outer dust window 90.
According to this embodiment, the dust window is tilted at an angle
.theta.1 relative to the plane 320 of the housing top surface. The
angle .theta.1 is generally between 7.degree. and 15.degree.. By
tilting the dust window 90 at a non-perpendicular angle,
reflections off its inner surface 322 are directed away from the
optical path of the image beam 84. Likewise, reflections off the
top surface 324 are deflected away from the scanned portion of the
sheet 60. The surfaces of the dust window can be provided with an
anti-reflection coating to further minimize bounce-back of image
light.
The image beam 84 strikes an elongated (in a direction
perpendicular to the page) reflector 92 located near a 45.degree.
angle to the incident image beam 84. The resulting reflected beam
326 strikes a second reflector 94 located at a slight angle
.theta.2 to the vertical plane 328 of the housing (the vertical
plane 328 being parallel with the incident image beam 84) the
slight angle .theta.2 causes the angled reflected beam 326 to be
transmitted perpendicularly to the plane 328 as a second reflected
beam 330. The second reflected beam strikes two 45.degree.
reflectors 96 and 98 to be re-reflected as a parallel beam 332. The
beam 332 passes through a focusing lens 100 that, in this
embodiment, comprises an F/6.2 objective lens. The lens 100
provides a focused, properly sized image at the plane defined by
the surface of the CCD camera element pixel array. As the image
moves relative to the camera, the CCD pixels eventually sample the
entire image (line-by-line).
The CCD camera element 101 is mounted on a support member 336 at
one end of the housing 86. Specifically, the CCD element 101
comprises the exposed photosensitive CCD pixel array 340 and a
transparent covering window 342 positioned in front of the array.
The pixel array 340 and window 342 are fixed together as a unit
according to this embodiment. Both the pixel array 340 and the
window 342 are arranged on the support member to be tilted at an
angle .theta.3 relative to a plane 338 passing perpendicularly
through the focused optical path 334. The angle .theta.3 is also
between 7.degree. and 15.degree.. In other words, the camera 340
and covering window 342 are disposed at a non-perpendicular angle
relative to the incident light striking them. By tilting the camera
and window into a non-perpendicular orientation relative to the
optical path, reflections of incident image light back into the
optical pathway are minimized. The angle does not substantially
affect the detected image, and hence, is acceptable for scanning
purposes.
It should again be noted that the reflectors, 92, 94, 96 and 98
each have a width sufficient to accommodate a full-width scannable
image. Since these reflectors are generally long and thin, they are
prone to experience strain that may distort the reflector shape and
degrade image quality. According to a preferred mounting procedure,
the reflectors are assembled as part of a base frame 360
constructed from a rigid material such as metal or plastic that
retains all components of the optical system. In particular, the
reflectors 92, 94, 96 and 98 are each mounted on a respective
mounting base 362, 364, 366 and 368 that each are, likewise,
secured to the base frame 360 by respective support member mounts
372, 374, 376 and 378. Each reflector is typically secured to its
mount by adhesive. The mounts are each secured to the base frame at
a predetermined mounting location by nuts, bolts or other
adjustable fasteners that enable adjustment of the reflectors to
optimize their positioning.
B. Camera System Enhancements and Output Signal Improvement
1. Dynamic Noise Suppression
By way of background, the dynamic range of an image is the measure
of the difference between highlights and shadows. It is defined as
a function of the density and the noise. The dynamic range (DR) is
the logarithm of the signal-to-noise ratio of the system. Deriving
its value in the logarithmic domain, it is represented as
DR=Density+Log(0.4343/image noise); where density is a "grayscale"
value ranging from 0.0 at maximum highlight and 4.0 near the shadow
limit; and the image noise is expressed as a root-mean-square (rms)
density noise value. For example, a density of 3.0 with an
associated noise level of 0.21 results in a dynamic range of
3.3.
When an image is processed by a CCD and associated circuitry, the
nature and quality of highlights can be controlled by regulating
the amount of light illuminating the image. However, the dynamic
range is significantly effected by noise in the shadow range. In a
conventional electronic imaging system, maximum light corresponds
to the maximum allowable electronic signal while absence of light
corresponds to the minimum electronic signal. There is an
approximately linear relationship between bright and dark as
exhibited by the output signal from the CCD. When a low-level
electronic signal is produced, the effect of background noise
becomes markedly more pronounced since the noise level is
approaching the output signal's normal amplitude.
FIG. 19 graphically represents a typical signal response curve 390
for an unmodified CCD imaging system for output signal voltage
versus shadow density (where 4.0 corresponds to a very dark
shadow). The largely linear response curve 390 exhibits a
significant and increasing noise range 392 once grayscale density
(e.g., shadows) exceeds approximately 2.0 density. The curve 390
forms a widening band of deviation 393 from the expected linear
response curve 394 (shown in phantom). This electronic noise is
exhibited by uncertain pixel output values.
Electronic noise in the system is a function of the rate at which
data is transmitted and the so-called "bandwidth" of the electronic
signal. The signal noise level (Vnoise) is proportional to the
mathematical square root of the bandwidth. The data rate is
typically fixed based upon system design requirements for a given
resolution. Conversely, bandwidth is variable for a given
resolution, and its attenuation can be used to control noise.
However, bandwidth is a function of system response and determines
how accurately the system will respond to a change in brightness of
the image. A narrower bandwidth, therefore, degrades the ability of
the system to respond to rapid changes in image brightness.
In general, the dynamic range of the system is limited by the noise
level of the signal. The magnitude of the noise level is a function
of the bandwidth and the signal varies as the square root of the
bandwidth. Hence, reducing the bandwidth of the signal reduces the
magnitude of the noise level. There is a practical limitation to
the amount which bandwidth can be reduced because the information
contained in the image signal is directly proportional to the
bandwidth. Therefore, there is a tradeoff between the required
information rate and the signal bandwidth.
The bandwidth of the output signal in the high intensity region is
largely fixed by scanner design and performance characteristics. An
example of a high intensity region detail in a developed X-ray
film, for which accurate reproduction is desired, is a thin,
hairline fracture appearing as a narrow dark line surrounded by the
white of the fractured bone. A wide bandwidth that facilitates
rapid response to contrast changes is, therefore, desirable for
such a region. In general, low-intensity (shadow) regions are less
dependent upon reproducing details with high contrast. For example,
shade changes in the dark background are irrelevant, and the
specific changes in tone within the line of the fracture do not
usually provide easily readable detail. Therefore, the bandwidth of
the signal in the low-intensity, dark regions can be narrowed
within certain parameters without degrading the image.
FIG. 20 illustrates a noise suppression circuit 400 that
dynamically adjusts the signal bandwidth of the system based upon
the content of the incoming image signal. This circuit 400 enables
the dynamic range of the signal (based upon the signal's Vmax and
Vmin) to be increased with a minimal effect on the underlying image
data since filtering of noise is applied in the shadow regions. The
circuit comprises a two-stage logarithmic (Log) amplifier having a
total dynamic range (DRLog) equaling 100 dB (e.g. a
minimum-to-maximum output ratio of 100,000). The Log amplifier is
divided into two cascaded stages denoted LOG1402 and LOG2404. It
is, however contemplated that separate signal processing amplifiers
can be utilized. The output of the Log amplifier first stage 402 is
connected to a variable low-pass filter. The filter 406 has an
adjustable filtering range of 10:1. The operation of the filter is
described further below. The filter 406 receives a control signal,
representing a filter cut-off value from a control electronics
block 408. The control electronics block comprises an amplifier
that responds to the input image signal 405, and by adjusting the
filter cut-off value based upon the prevailing characteristics
(highlights or shadows) of the signal 405. The output of the filter
406 is connected to the Log amplifier second stage 404. The
filtered output of the Log amplifier second stage 404 and the
output of the Log amplifier first stage 402 are combined at a
summing circuit 410 to produce an output Log analog image signal.
Each Log amplifier stage 402 and 404 processes one-half (50 dB) of
the total dynamic range of the input image signal 405. For a 1.0
Volt maximum signal level, the first stage 402 processes signals in
the range of 0.00316 Volt to 1.0 Volt and the second stage 404
processes signals in the range of 0.00001 Volt to 0.00316 Volt. The
first stage and second stage Log outputs are summed to provide the
total 100 dB Log amplification range.
As discussed above, variable low-pass filter 406, connected between
the two Log amplifier stages 402, 404, is regulated by the control
electronics block 408 to vary the bandwidth of Log amplified
low-level signals. The filter is adjustable over a 10:1 range,
which corresponds to a reduction in electronic noise by 3.16 times.
This reduction translates into an increase in the dynamic range of
10 dB. According to this embodiment, the amplifier function of the
control electronics 408 adjusts the low-pass filter 406 so that the
filter allows a maximum bandwidth to pass when the input signal is
greater than 0.002 Volt. When the signal is less than 0.0003 Volt,
the control electronics 408 adjusts the filter to allow the minimum
bandwidth to pass. The bandwidth filtering response between these
limits is adjusted by the control electronics based upon a
monotonic response function. The values for maximum and minimum
bandwidth can be established based upon a desired signal-to-noise
ratio that is acceptable for shadow images. An acceptable level can
be established based upon experimental data, varying the bandwidth
level and observing the quality of the image or based upon
previously established industry standards for image quality.
The signal output from the summing circuit 410 is a Log-amplified
analog image signal 411 having a 100 dB range. For the signal to be
translated into a storable data file, it is converted into a
digital signal by a conventional analog-to-digital converter 412.
The converter 412 can be part of the noise suppression circuit 400,
or can be located at another point along the signal path.
It is common in certain fields, such as the medical arts, to store
and display image data in logarithmic form. The contrasts between
details are enhanced during display when image intensities are
reproduced based upon a logarithmic translation. In one embodiment,
an anti-log function is used to translate the logarithmic digital
signal output from the converter 412 into a linear digital signal
for file storage as a non-logarithmic image file. The anti-log
function can be implemented as a discrete circuit or can reside as
a software procedure within the CPU 52. The anti-log function,
according to one embodiment can comprise a look-up table in which
the logarithmic image data is compared to linear data values, and a
translation of data is made based upon the compared values. Note
that the anti-log function according to this embodiment occurs in
the digital signal domain. It is contemplated that the anti-log
function can be implemented in the analog domain, with the signal
subsequently translated into a digital format. Note, as an
alternative, scanning can be implemented fully in the linear signal
mode, in which case an anti-log function is not employed and
filtering occurs only in the linear domain.
While base-ten Log amplifiers are used to process the signal
according to this embodiment, it is contemplated that other forms
of signal amplification can be employed. Appropriate
deamplification functions are used to return the signal to a
storable linear format. Such signal processing devices, shall be
termed signal "amplification circuits" herein.
FIG. 21 shows an exemplary response curve 420 for signal voltage
versus image density obtained using the dynamic noise suppression
processes described above. The curve 420 exhibits reduced noise
characteristics (422) in the shadow regions. The band of deviation
423 is markedly narrower than the unfiltered band (393 in FIG. 19),
and the signal, generally, deviates less from the expected linear
response curve 395 (shown in phantom). As illustrated, noise
associated with shadow details is effectively reduced without
sacrificing needed bandwidth in the high-intensity regions of the
image.
2. Pixel Intensity Averaging
The CCD element of this embodiment is arranged to provide a
resolution of 584 pixels per inch (PPI) for a 14-inch wide image.
Medical X-rays and other films often permit a lower resolution--on
the order of 75-150 PPI. Using a higher-than-necessary resolution
entails a waste of data storage space to store extra image data and
slows the scanning and storage processes by causing the various
system processors to operate on a larger volume of data. It can
also make transmission of image data over telephone lines or other
data transfer networks prohibitively time-consuming.
FIG. 22 schematically depicts a linear array 450 of picture
elements (CCD "pixels") within the CCD element. The CCD pixels
arranged in the direction of the scanner width, or "fast scan"
direction (arrow 452), and are denoted P1, P2, P3, P4, P5, P6, P7,
P8 and P9 according to this example. In all, approximately 8,000
CCD pixels are provided in the linear array 450 according to a
preferred embodiment of this invention. While not shown, this full
range of pixels can be denoted as P1-P8000 for the purposes of this
discussion. As described previously, the pixels view discrete lines
of the image spaced approximately one pixel width apart in the
feeding or "slow scan" direction (arrow 454). In other words, every
8 milliseconds, the array locks in a line image. According to this
embodiment, the motor 46 is instructed to move 152 steps/second to
attain 584 PPI resolution. The movement of the sheet in the slow
scan direction is sufficient to present the next pixel width
(approximately 0.00171 inch of the full-size image) to the array. A
series of adjacent scan lines along the length of the image are
represented by contiguous rows of pixels P1-P9 denoted as S1, S2,
S3, S4, S5, S6, S7 and S8. Each row is a representation of the
scanning of a different, adjacent part of the image by the same set
of pixel--separated by time and space.
A method for reducing the camera's rated resolution, and smoothing
the response curve involves the summing of the intensities of
adjacent pixels in both the fast scan and the slow scan direction,
and deriving an average intensity value for the group. The space
occupied by the pixels in the image data field is provided with
this average value and the scanned image appears as a single large
pixel occupying the space of the averaged pixels displaying the
average intensity value. 146 PP1 resolution can be obtained by
dividing the native resolution 584 PP1 by a predetermined divisor.
In the fast scan direction, groupings 456 of two pixels P1 and P2,
P3 and P4, P5 and P6, and P7 and P8 can be made. Similarly
groupings 258 of four pixels P1-P4 and P5-P8. Additionally a large
grouping 460 of eight pixels P1-P8 can be made. Groupings of
adjacent pixels across the entire length of the array 450 are, thus
made. Depending upon whether 2, 4 or 8 adjacent pixel intensities
are averaged, the effective number of CCD pixels is reduced to
approximately 4,000, 2,000 or 1,000, respectively.
FIG. 23 details a specific averaging procedure for adjacent pixels
oriented in the fast scan direction. First, the binary value of the
intensity derived from each of these pixels is measured in step 462
by the CPU or an associated averaging circuit. The CPU or circuit
is provided with instructions to average either 2, 4 or 8 pixels
depending upon the desired reduction in resolution. These binary
intensity values are then added in step 464. Having a total binary
intensity value for a group of 2, 4, or 8 continuous pixels, an
average can be derived by shifting the binary sum by 1, 2 or 3 bits
in step 466. Whether the sum is shifted by 1, 2 or 3 bits depends
upon whether 2, 4 or 8 continuous pixels have been summed
respectively. The average value is then transmitted downstream for
further data processing and storage in the digital domain in step
468. Each adjacent group of 2, 4 or 8 continuous pixels is read in
the fast scan direction in step 470. The averaging of adjacent
array pixels continues for each scanned row (S1-S8, and so forth)
until the entire image has been scanned in the slow scan
direction.
With reference again to FIG. 22, the rows of slow scan pixels S1-S8
are also averaged. In other words, P1 can be intensity-averaged
over a plurality of adjacent scan lines S1-S8. Likewise P2-P9 can
each be intensity-averaged over a plurality of contiguous scan
lines. The signals derived from adjacent scan lines can be grouped
similarly to the grouping of adjacent array pixels for a single
line as described above. For example, groupings 480 of two lines S1
and S2, S3 and S4, S5 and S6, and S6 and S7 can be made. Similarly
groupings 482 of four lines S1-S4 and S5-S8 or groupings 484 of
eight lines S1-S8 can be made.
FIG. 24 details a basic pixel intensity averaging procedure for
line data in the slow scan direction The measured intensity of the
output signal from a group of 2, 4 or 8 adjacent scan lines (the
aggregate signal of all pixels in the array) is averaged in step
486. The averaging of lines is accomplished by increasing the scan
speed so that the CCD views more than one line within the allotted
line scan duration of 8 milliseconds. The line of CCD pixels
essentially views a "blur" of several discrete intensities in the
slow scan direction. The approximate average intensity of the
passing group is transmitted by each CCD pixel. The speed of the
drive motor 46 is increased 2, 4 or 8 times, to attain the desired
average of 2, 4 or 8 lines, respectively. In step 488 the next
group of 2, 4 or 8 contiguous lines is then scanned at the desired
scan speed to derive an average slow scan intensity value. The
process continues until the entire image has passed through the
scanner.
Fast scan averaging and slow scan averaging are employed
simultaneously. For example, two adjacent fast scan pixels P1 and
P2 are averaged, and two lines S1 and S2 are scanned in one scan
interval. The result is a grouping of four total image pixels into
one larger square image pixel. The corresponding resolution is
reduced four times, generating a desired 146 PPI image. Likewise,
when four fast scan pixels are averaged, a group of four slow scan
lines are scanned in one scan interval. The resulting averaged
group of image pixels comprises a total of sixteen pixels in a
square with a single averaged intensity. The corresponding area
resolution is reduced sixteen times. However, it is accepted in the
art to refer to resolution only in the linear, fast scan direction,
even though resolution is typically reduced in both directions to
maintain proper scale. Hence, by averaging eight pixels in each of
the fast and slow scan directions, the resulting fast scan linear
resolution is reduced from 584 PPI to 146 PPI. It is expressly
contemplated that any number of pixels (e.g. 3, 5, 6 or 7) can be
averaged in the fast scan and slow scan direction to generate
images having other desired resolution. 2, 4 and 8 pixel averages
are made in a preferred embodiment to enable binary division of the
fast scan intensity value as described above. Averaging can occur
in either order (e.g. fast scan first and slow scan second, or vice
versa). Sums for pixel values in one direction are then averaged in
the second direction to obtain the final average for the square in
both orthogonal directions.
The averaging of adjacent fast scan pixels achieves an improvement
in the signal-to-noise ratio of the CCD element that extends its
dynamic range significantly and that produces a smoother-appearing
image. Averaging two adjacent pixels decreases the effective
electrical noise by 2 while extending the dynamic range by Log(2),
equal to 0.15 grayscale density. Taking the slow scan resolution
reduction into account, the noise is in fact reduced by 4 or two
times, while the extension of the dynamic range overall is doubled
to 0.3. IV. Image Size Determination
A. Size Measurement and Associated Scan Adjustment
If the size of a sheet is not accurately known by the scanner, the
camera element may scan beyond the lead and trail edges and beyond
the widthwise edges, producing image data for the open area around
the sheet as well as the image. Since image data takes up
substantial quantities of data storage space and processing time,
it is desirable to omit as much external scan data as possible,
particularly in the fast scan direction, where the widthwise edges
may be much narrower than the field of view.
Described herein is a image size-measurement procedure particularly
suitable for developed X-ray film such as the film sheet 60 shown
in FIG. 2. The film sheet 60 includes a predictably sized,
substantially rectilinear outer perimeter 74. At least part of the
background 68 adjacent the perimeter 74 exhibits a substantially
dark shade providing a significant contrast with the unattenuated
free space outwardly of the perimeter. In any case, the base film
exhibits an increased density relative to the surrounding,
unattenuated free space.
According to the procedure detailed in FIG. 25, the sheet 60 is
input to the feed rollers pairs 38 and 40. As the lead edge 501 of
the sheet 60 passes into the rollers, the CPU 52 activates the
scanner's illuminator and CCD camera in step 500. The scanner CPU
does not create an image data file for storage by the microcomputer
107 at this time. Rather the CPU polls the CCD for the output
signal in the fast scan direction continuously until the output
changes from the white shade exhibited by the unblocked
illumination light to shadows of the background in step 502. Since
the sheet's width is not yet known, the procedure polls for an
intensity transition small segment of the lead edge located, for
example, in the relative center of the field of view. This
indicates the location of the lead edge 501. Since the sheet edges
may be skewed slightly, the procedure can be instructed to poll for
the light/dark margin transition line over a few slow scan lines.
If the transition occurs within a predetermined number of lines,
then the top margin is confirmed. The location of the lead edge is
compared to the relative position of the motor 46. If a stepper
motor is used, the "steps" can be counted. Other types of drive
motor can be provided with an encoder or other data acquisition
device for measuring rotation that is interconnected with the CPU.
The CPU, thus, monitors the motion of the motor 46 as it feeds the
sheet 60 into the scanner. The relative position of the motor in
the slow scan direction is, hence, stored by the CPU as indicating
of the location of the lead edge 501 (FIG. 2).
In step 506, the CPU then polls the entire fast scan direction for
the transition between the side edges 510 and side background lines
512 (FIG. 2). The location of the side margin lines is derived
based upon the location in the CCD pixels at the boundaries between
a maximum highlight and a darker attenuated light. The location of
the widthwise boundaries can be scanned continuously as the sheet
progresses through the scanner, mapping the boundary locations to
the location of the motor 46. Alternatively, the CPU can poll the
widthwise boundary at one or more specific slow scan location(s)
and establish a scan width for the entire image based upon the
measured value(s).
In step 514, the CPU continues to poll for a transition at the
trail edge 515. Again, the entire fast scan direction can be
scanned, or the scan for the trail edge transition can be limited
to a particular widthwise segment that is less than the total fast
scan field of view. The location of the trail edge 515 is located,
and before the sheet 60 is ejected by the downstream feed rollers
40, the motor is reversed in step 516. The trail edge sensor can be
used to signal the CPU as to the relative location of the trail
edge, so that the sheet is reversed before the trail edge exits the
rollers 40. The sheet is driven in reverse, first by the rollers 40
and, thence by the rollers 38, to a start position in which the
lead edge 504 is again upstream of the camera's field of view. The
CPU now has a data file mapping the locations of the sheet edges
versus the relative rotational position of the motor (based upon
the motor's step locations).
Knowing the relative rotational position of the motor at the end of
the first scan cycle, the CPU restarts the feeding process in step
518. The sheet is scanned with the camera started at the known
location of the lead edge. Image data is acquired and transmitted
by the camera based upon the recorded widthwise margins, data
scanned outside of the edges is neither stored nor transmitted to
the microcomputer 107, saving data processing time. When the known
length has passed through the field of view of the camera, the
camera is deactivated in step 520. The sheet 60 is then ejected by
the rollers from the scanner in step 522. The next sheet is fed
into the scanner in step 524 and the procedure above is
repeated.
The CPU can be provided with the approximate expected size for the
scanned images as a default value. In the case of developed X-ray
film, the sizes tend to be standardized, and a data input key
indicating relative sheet size can be provided to the scanner
according to an embodiment of this invention.
As noted above, the size measurement procedure described above is
applicable to a sheet having a minimal intensity transition at its
edges. A sheet having highlights at its edges will still attenuate
light intensity to a small extent. The transition to the lower
intensity value can be used to establish the edges of the
sheet.
B. Illumination Field Size Adjustment
The techniques described above for controlling the scan size also
enable selective control of the field of the illumination assembly
according to an embodiment of this invention. An advantage to
providing an illuminator comprised of a plurality of discrete light
sources is that the light sources can be selectively deactivated
where light is unneeded. It is desirable to attenuate unneeded
light outside of the image field, since this light can fall into
the camera assembly as stray light which can degrade the
photometric accuracy of the image. By locating the widthwise limits
of the image, LEDs that project light beyond the limits are
deactivated by the CPU. FIG. 26 shows a flow diagram for an
illumination control procedure according to this embodiment. First,
the CPU identifies the widthwise (fast scan) edges or other limits
of the image 550. This can be accomplished using the automatic size
sensing procedures described above, by use of movable edge guides
(not shown) on the stack tray having width sensors that signal the
CPU or by manually entering the width of the document into the CPU
through the microcomputer's user interface or another data entry
device interconnected with the CPU. In a preferred embodiment, it
is contemplated that a document is fed through the scanner in a
centered orientation relative to the width. Thus, a sheet or image
that is narrower than the maximum width will have an equal width of
free space outward of each widthwise edge.
After identifying the widthwise limits of the image in step 550,
the CPU associates the locations of the widthwise limits of the
image with the physical locations of particular LEDs in step 552.
In other words, LEDs that are located outwardly of the limits of
the image are identified, based upon their known locations within
the array, and the LED# of each identified LED (e.g. LED1, LED2 . .
. and . . . LED58, LED59 and LED60) is stored in a file that
denotes unneeded LEDs in the array for the particular scanning
operation currently underway. The LEDs can be associated with the
widthwise edges in a variety of ways. The size of the sheet can be
provided to a size equation or look-up table that correlates
particular LEDs in the array with a given size value. As size
increases, the number of unneeded LEDs on each edge is reduced.
Alternately, the CCD camera element can be used to directly
determine which LEDs are registering a free space intensity. Since
LEDs are mapped to CCD pixels during the calibration process, the
LEDs can be readily identified by which CCD pixels are registering
full intensity.
In step 554 the LEDs are deactivated by transmitting a signal
representative of a minimum driving current or no current the each
identified LEDs' programmable current source via the data line 162
(FIG. 7). Other LEDs in the array are driven at their standard
driving current based upon each LEDs' respective calibration value.
It is contemplated that one or more LEDs that are adjacent, but
outward of, to the limits of the image can be driven. This can
ensure that the edges are fully illuminated since some cross talk
between LED light patterns is known to exist.
Once the scan using the reduced-width LED array is completed, the
procedure returns to begin the next scan in step 556. The LED array
can operate using the same adjusted width if the CPU is instructed
to do so, or a new set of width parameters can be selected. The
procedure is then repeated using the new width parameters.
As described below, it is contemplated that portions of the image
within the overall boundaries of the sheet or image can be scanned
individually. Such "regions of interest" can have widths that are
less than the overall width of the image. By specifying the width
using, for example, the procedures described below, the LED array
can be adjusted to deactivate LEDs falling outside of the width of
the region of interest. Note that the region of interest need not
be centered with respect to the overall image. By providing
appropriate coordinates referenced, for example, to the scanner
centerline or a widthwise edge, the corresponding unneeded LEDs can
be identified. Likewise it is contemplated that two or more
separate LED array widths can be used in a single image. The
location along the length of the document is tracked, using the
drive motor as a reference, and the width is changes as
predetermined lengthwise locations are reached.
It is further contemplated that sheets can be fed registered to an
edge of the tray rather than the centerline. Appropriate edge
guides (not shown) can be provided to facilitate side registration.
In such an arrangement, the LEDs on a single side are is generally
deactivated based upon known image size and/or edge location data.
V. High-Resolution Overlays
As discussed generally above with reference to pixel averaging
procedures, it is often desirable to reduce the resolution of the
scanned image to generate a smaller image data file and speed the
scanning process. The current standard for resolution radiological
scanning is approximately 73-146 PPI, which enables an image to be
viewed on a monitor having between 1-2K of horizontal pixels.
Reduction of the scanned resolution from the higher native
resolution of the CCD element to this standard is employed
according to an embodiment of this invention. This lower resolution
is also advantageous when transmitting images over a computer
network or facsimile machines serviced by voice-carrying telephone
lines since the data transmission speed of theses carriers is often
highly limited.
A higher resolution image of a particular region of interest on an
overall image is sometimes required for specific diagnostic
purposes. For example, an image of a broken bone may reveal a
complicated fracture pattern requiring closer study. A
low-resolution image may be insufficient to accentuate important
details of the fracture. However, scanning an entire image at
high-resolution simply to capture a relatively small region of
interest wastes significant data storage and scanning/processing
time.
In general, medical images can be formatted according to the
well-known DICOM-3 standard, promulgated by the American National
Standards Institute (ANSI). The DICOM-3 standard defines a data
structure that enables related images to be grouped together to
form an image "series." As described above, diagnostic X-rays are
usually taken as a series, such as left, right, top and bottom
views of an injured area. Views of the same area at different times
can also comprise a series, as can any other associated group of
images. Using the functionality of the DICOM-3 standard, it is
possible to generate associations between a parent image and a
higher-resolution region of interest, "child" image.
FIGS. 26 and 27 respectively illustrate a display and procedure for
generating an associated high-resolution child image of a region of
interest in the lower resolution parent image. The monitor screen
600 projects a low-resolution (approximately 146 PPI) image 602
having been recently scanned from a developed X-ray film sheet (not
shown). The main subject 604 of the image 602 contains a region of
interest 606 that is much smaller than the subject 604 and overall
image 602. The screen in this embodiment is set up in a Microsoft
Windows.RTM.-style graphical user interface format. The user
interface is controlled based upon the microcomputer keyboard 111
and mouse 113 (FIG. 1) that enable entry and manipulation of text
data and control of a screen cursor 608. A button bar 610,
manipulated by the cursor 608, is provided to act upon screen data
and enter instructions. General control of the scanning process can
be undertaken using the button bar 610 and associated data entry
devices.
The parent image is provided with a file identifier 612, that is
printed adjacent the image 602 in this embodiment. The identifier
612 can also be positioned within the background or margin of the
projected image or at another acceptable location. The image can
include various size notations 614 and 616 in the respective
horizontal (widthwise) and vertical (lengthwise) directions. Other
notations can also be provided as appropriate to assist the reader
in understanding the nature and content of the image, such as
information related to patient name, time, date, location, series,
diagnosis and miscellaneous notes.
The main subject 604 of the image 602 contains a region of interest
606 delineated by a dashed-line border 618. The border can be
visible on the screen or can be represented by its comers A, B, C
and D. The shape of the border can be varied depending upon the
geometry of the region of interest. The border 618 is defined by
movement of the cursor in conjunction with commands entered via the
button bar 610 or keyboard (or both). Information regarding border
coordinates is shown, in this embodiment in a window 619. A variety
of conventional applications are available for defining borders on
screen images. A separate identifier 620 is printed within the
border. Like the parent image identifier 612, the region of
interest identifier 620 can be located at any acceptable location
that is understandable and readable. The identifier 620, in
substance, provides an annotation to an associated file that
contains a higher resolution rendering of the delineated region of
interest. That rendering is shown on the screen 600 as a window
622. The window 622 contains an identifier 624 that corresponds to
the region of interest identifier 622 annotating the parent image
602, the nature of the identifiers is based upon user preference.
Related identifiers can be designated as different series numbers
or as sub-files to specific images in a series.
The particular screen representation and information shown in FIG.
26 is only one possible display format. The DICOM-3 standard
relates to a format for storing data that enables linking of image
files to create "overlays" wherein one image file includes an
annotation that links that file to another file. As described
herein, the overlay is an identifier for a separate high-resolution
child image of a portion of the parent image. The DICOM-3 standard
makes possible the illustrated display of files, and a variety of
other organizational structures known to those of ordinary skill.
For example, files can be transmitted via DICOM-3 over a
telecommunications link. The advantage to the linking procedures as
described herein is that a user has instant access to associated
data in linked files, and these files will not become lost or
inadvertently separated. Using the above-described format, minimum
storage and communications resources are employed since high
resolution data is used only where needed. Note that several
regions of interest can be linked to a single file according to
this invention. Each would utilize its own annotation on the parent
image.
A basic procedure for establishing a high-resolution overlay is now
provided, with reference to the flow diagram in FIG. 27. Initially,
a film is scanned and the microcomputer and CPU are instructed to
format image data according to the DICOM-3 standard in step 650.
The entire image is stored in low-resolution format using the
procedures described above. The CPU tracks the movement of the
drive motor to monitor the relative locations of portions of the
scanned image in one embodiment.
The low-resolution image is displayed on the monitor screen once
scanning has been completed in step 652. An operator can identify a
region of interest on the low-resolution monitor display. Using the
mouse to manipulate the cursor, and the button is bars, the user
can delineate borders for the region of interest. The borders are
internally processed by the microcomputer and the CPU to determine
the scanning location on the film of the region of interest. This
location will be used to perform a rescanning of the region of
interest as described further below. Before rescanning, the user or
the computer assigns an annotating identifier to the region of
interest in step 654. This identifier is appended to the parent
image file according to the DICOM-3 standard, and appears as an
overlay annotation on the parent image.
The scanner can either eject the scanned film, or reverse it to its
original starting position within the scanner in a manner similar
to the size-sensing embodiment described above. In step 656 the
film is then rescanned while the CPU polls for the region of
interest. The CPU uses the position of the motor to find the
slow-scan location of the image and uses fast-scan location data
derived from the original low-resolution image to perform a
high-resolution scan and storage of data in the region of interest.
Other image data, outside the region of interest is not stored. The
mechanics of the high-resolution scan are similar to those
described with respect to the size sensing embodiment described
above.
In step 658 the high-resolution data is stored in a file according
to the DICOM-3 format and is identified by a link that corresponds
to the annotation on the parent image. In further transmissions or
data transfers, the resulting parent and child files are provided
as a unit unless otherwise instructed. The annotation of the parent
image with the identifier can include the provision of hypertext
command structures according to known techniques. According to such
a command structure, when a user manipulates the identifier with
the cursor, it causes the child image to appear in a separate
window or a new screen view. Finally, in step 660, the next image
is scanned by an operator.
It is contemplated that several high-resolution scans can occur in
a single rescanning pass. Each region of interest is located on the
parent image as it is rescanned based upon the procedures described
immediately above. It is further contemplated that a plurality of
rescanning passes can be made to add further regions of interest to
the parent image file. Alternatively, if the location of the region
of interest is accurately known, the CPU can be instructed to
perform a high resolution scan of the known area based upon
previously entered film sheet coordinates. Accurate registration of
the sheet is desirable. Registration can be determined by carefully
feeding the sheet, or by performing a prior size-sensing scan as
described above. VII. Camera Assembly Bias Calibration
As discussed above, the individual pixels of the CCD camera element
each output individual intensity value signals that may vary
significantly from pixel-to-pixel, even in the presence of a
substantially constant incident light intensity. This variability
generally results from slight manufacturing differences between
pixels in the CCD array and nonuniformity in the illumination
assembly. The differences between output signals are minimized, in
part, by providing appropriate correction factors to each CCD
pixel's output signal. In particular, each CCD pixel may provide a
different output signal (dark current) for the same
shadow/dark-viewed intensity. Additionally, each CCD pixel may
exhibit a different gain in response to illumination. Gain can be
determined by comparing the dark calibration and light calibration
output signal value for a pixel. Light and dark calibration of the
CCD camera element are described generally above with reference to
FIG. 8.
According to a preferred embodiment it is contemplated that the
gain of all CCD pixels is initially adjusted as a group during the
illumination calibration step. The bias adjustment procedure (now
to be described) is then initiated. Following the bias adjustment,
a fine-tuning of the gain of each individual CCD pixel can also be
performed. The initial and fine-adjustment of gain can be
conventional and, therefore, is not described in detail.
FIG. 29 details the output stage of a typical CCD pixel of the CCD
camera array according to a preferred embodiment of this invention.
The CCD pixel 670 converts incident light 672 into an output linear
analog intensity signal that, as described above, is sampled every
8 milliseconds. The linear output signal can be characterized as
either a variable voltage or a variable current signal based upon
Ohm's Law. For the purposes of this discussion, both a voltage and
a current characterization are used. The linear output signal is
transmitted to a linear preamplifier 674 that increases the range
of the linear intensity signal to a desired scale, such as 1 volt.
The amplified linear signal is transmitted from the linear
amplifier into a base-10 logarithmic amplifier 676. The logarithmic
amplifier 676 can comprise the two-stage log amplifier detailed in
FIG. 20 with reference to dynamic noise suppression. It is
contemplated that the analog image signal of FIG. 20 can be the
signal output from the linear preamplifier 674, and that the
logarithmic amplifier 676 can, in fact, include all amplification
and filtering elements for providing dynamic noise suppression
according to FIG. 20. The logarithmic amplifier stage outputs a
logarithmically expanded analog signal that is transmitted to an
analog/digital converter 678. The analog/digital converter in this
embodiment generates a corresponding 12-bit digital numerical
intensity value in the logarithmic domain for the corresponding
input logarithmic analog signal. This digital signal is passed to
the CPU as system response data.
A digital/analog converter 680 transmits, via a matching resistor
682, a bias current to the linear preamplifier 674 that produces a
summed current input to the logarthimic amplifier. The CPU provides
numerical digital bias values that are converted into corresponding
analog voltage/current values by the digital/analog converter 680.
The bias voltage/current is summed at the linear preamplifier 674
to generate an output signal that comprises a bias-adjusted,
amplified linear signal. This bias-adjusted, amplified signal is
input to the logarithmic amplifier 676, and finally output in
logarithmic-digital form to the CPU. In subsequent scan operations,
the output of each CCD pixel is continuously summed with an
assigned bias-adjustment factor, and this bias-adjusted intensity
value is stored as image data.
A procedure for calibrating the CCD array to adjust the bias of
each pixel is detailed in FIG. 30. To exemplify the bias adjustment
procedure reference is also made to a series of graphs shown,
respectively, in FIGS. 31, 32 and 33.
The bias adjustment procedure begins in step 700 in which the
illumination assembly is deactivated so that a maximum dark
intensity is presented to the CCD array. A group of contiguous
pixels within the CCD array is preselected to provide test values.
Adjustment of bias for each pixel in the overall CCD array is
subsequently based upon factors initially derived from this group.
According to one embodiment, a group of approximately 4,000 pixels
is selected. This is approximately one half the total number of
pixels in the CCD array.
The pixels of commercially available CCD arrays exhibit a
relatively uniform pixel-to-pixel output. A minimum number of
pixels may exibit an output variation, typically in the range of
one decade in the logarithmic domain. It is acceptable to establish
a bias response curve based upon a relatively large group of CCD
pixels since only a small number of pixels will vary significantly
from a desired norm. The small number that show marked variablity
are easily smoothed over using statistical smoothing and averaging
techniques. The uniformity of dark current output across the array,
in fact makes possible the efficient bias adjustment procedure that
is now being described.
The output of the first pixel in the group is summed with a first
bias value at the linear preamplifier stage at step 702. In this
embodiment, the 12-bit digital numerical value for each bias is in
the range of 2,043-6,149, which is chosen arbitrarily in this
embodiment. This represents 4,096 different bias states centered
relative to 4,095. The first bias generates an associated negative
voltage/current that produces a negative system response data point
721 (see curve 720 in FIG. 31) from the logarithmic amplifier 676.
The logarithmic amplifier 676 is precalibrated so that the bias
voltage/current corresponding to the test bias 2,043 produces a
negative system response at the output of the logarithmic
amplifier. The analog/digital converter 678 at the output of the
logarithmic amplifier generates an associated system response
digital value shown as -2,700. Note that the negative system
response is a convention used to effect calibration as described
below, since the laws of mathematics do not permit negative
logarithms in a literal sense. The logarithmic output scale on both
sides of the zero-system response point 723 (FIG. 31) is intended
to be represented by the response curve 720. The first digital
system response value (-2,700) is read in step 704 by the CPU, and
stored in the CPU memory with the corresponding test bias 2,043 as
a data point in step 706.
The procedure queries, in decision step 708, whether a last pixel
has been read. Since the first pixel is not the last to be read,
the step 708 branches to the bias incrementing step 710. The
digital/analog converter 680 is set up to provide a range of
incremental current/voltage bias inputs to the logarithmic
amplifier 674 that are summed with the associated dark current
output for the linear preamplifier 674. As described generally
above, the linear output of each pixel in the group is summed (at
the preamplifier stage) with a different, upwardly ramped bias
value to generate an overall system response versus bias curve 720
as defined in FIG. 31 in step 710. Each new system response is read
in step 704, and stored as a digital, numerical value in step
706.
The procedure continues to upwardly increment the bias values,
summing an associated current/voltage increment to the linear
output of each new pixel until the last pixel output in the group
has been summed with the last bias value. The corresponding
voltage/current values pass from negative to positive at digital
bias value 4,095 (in the iluustrated example). A system response of
approximately 0 occurs at the cross-over point. The system response
from bias values 4,096 to 6,149 produces associated positive
voltage/current increments at the logarithmic amplifier 676. The
calibration of the logarithmic amplifier produces a corresponding
positive curve segment 724 with a logarithmic response scale equal
to the negative segment 722. For a test bias having a digital value
of 6,149, the digital system response is +2,700. This range of
system responses output from the logarithmic amplifier (-2,700 to
+2,700) represents the probable range of outputs transmitted from
the logarithmic amplifier 676 when a given group of CCD pixels are
each exposed to the same dark intensity. Note that by assigning a
different ramped bias to each pixel, rather than assigning the
ramped values to the same pixel(s), the entire response curve can
be derived in a single 8-millisecond scan cycle.
When all pixels in the group have had bias applied, and have had
the applied bias read as an output system response, the decision
step 708 then branches to the curve-derivation step 730. The curve
720 of FIG. 31 is representative of a map of system response versus
test bias in the digital (numerical) domain. In deriving the curve,
the number of data points in the map can be reduced from the
approximately 4,000 originally collected. For example, every eight
contiguous data points can be averaged to derive a resultant mean
value for test bias and a mean bias for system response. In one
embodiment, every eight data points are averaged. Other
conventional smoothing functions, such as least-squares can be
applied to the curve 720 in step 732. Curve-smoothing is generally
desired since each pixel will generate a slightly different system
output to a given dark intensity that will cause noise in the
resulting system response curve. In addition, certain pixels may
exhibit significant variability, which statistical smoothing helps
to eliminate. With appropriate statistical averaging, however, the
overall output response of the pixels over given a range of bias
values should define a substantially continuous curve.
In step 734, the curve 720 (FIG. 31) is inverted, by translating
the data points to produce a curve 736 (FIG. 32) of test bias
versus system response. By way of example, the curve 720 shows that
a digital numerical response of 2,700 is obtained from a 5,600
digital numerical bias. Likewise a response of 1,800 is obtained
from a 4,608 bias. The data points for a given system response are
generated using associated test biases as detailed in FIG. 31. In
other words, it can be determined that a system response of 2,700
corresponds to a test bias of 5,600, and that a system response of
1,800 corresponds to a bias of 4,608.
Based upon the data points of the inverted curve 736 (FIG. 32), a
final table of adjustment factors is derived in step 738. This
table is represented by the curve 738 in FIG. 33. Each system
response has associated therewith a test bias as illustrated by the
curve 736. In step 738, the procedure determines what adjustment
bias in the linear domain must be applied to bring the output of a
pixel to an approximate digital numerical value of 0 (or any other
desired dark current response) when the actual digital output of
the pixel in the logarithmic domain deviates from the desired 0
response. For example, a bias of 5,600 is 1,505 counts higher than
the bias that produces a 0 response. In this example the bias that
produces a 0 response is 4095. Note that a bias of 5,600 is
associated with a system response of 2,700. The difference between
the test bias of 5,600 and the dark current value of 4,095 is
1,505. This difference is calculated for each desired data point.
The resulting difference 1,505 is then subtracted from the dark
current value of 4,095 to obtain an adjustment factor of 2,590. The
adjustment factor is then mapped to the associated system response
of 2,700. Hence, if a pixel has an output of 2,700, the associated
bias adjustment factor 2,590 will restore it to a system response
of 0. Likewise, an output of 1,800 requires an adjustment bias of
3,582 to obtain a 0 response value from the logarithmic amplifier
676. Similarly, a logarithmic output of -2,700 corresponds to a
bias adjustment of 6,149 at the linear stage to raise the dark
current output of a CCD pixel in the logarithmic domain to 0. Of
course, if the pixel provides an unadjusted logarithmic domain
system response of 0, then a bias of 4095 is summed with the output
at the linear output stage.
As described above, to speed the bias calibration process the curve
730 can be formed using data points that are spaced-apart by
increments greater than 1. For example, incremental CCD pixel
output values separated by 8 points can be stored in a map with
their associated correction factors. This reduces the size of the
file, speeding calibration. Adjustment factors for CCD pixel offset
values falling between the increments can be derived by taking the
correction factors for the nearest offset values on either side of
the read value and performing an interpolation. Conventional linear
interpolation can be employed so long as the increments are
sufficiently close to each other. According to one embodiment,
eight points are averaged into a single data point having a mean
value for offset and adjustment factor. This averaging serves to
smooth the adjustment factor curve 738 in addition to reducing the
size of the adjustment factor table.
Finally, having derived an adjustment factor curve/table in step
738, the logarithmic output value of every CCD pixel in the array
is read during exposure to the same dark intensity, and a table of
bias adjustment factors is built. In step 740 an appropriate
correction factor (derived from curve 738) is individually mapped
to each CCD pixel based upon the amount that the pixel's
unadjusted, logarithmic domain output deviates from the established
baseline 0 response value. These correction factors are appended to
the linear preamplifier output of each CCD pixel, respectively,
whenever that pixel transmits an intensity signal.
Note that the values described above have been represented as
base-ten counterparts of digital integer values. Appropriate
analog/digital and digital/analog converters are used at the bias
input and logarithmic output to produce these digital values. Such
value are used by the CPU in manipulating data. It is contemplated
that the logarithmic and linear values can be expressed in
voltage/current terms, and it is recognized that the digital values
described herein are representative of underlying voltage/current
values. Furthermore, while a logarithmic amplifier is used to
process the linear output, any form of signal processing circuit
can be employed, and the resulting signal can exhibit any form of
linear or non-linear characteristic. Accordingly, the term "signal
converter," as used herein shall be taken to denote any type of
signal processing circuit that produces a characteristic linear or
non-linear output signal from a raw CCD pixel linear output signal.
Note also that a second gain calibration can be applied to the
pixels, following the above-described bias calibration procedure,
according to an embodiment of the invention.
The foregoing has been a detailed description of preferred
embodiments of the invention. Various modifications and additions
can be made without departing from the spirit and scope of the
invention. For example, the illumination assembly of this invention
can be applied to a variety of scanner systems either reflecting
illumination light from or transmitting light through the image.
The camera assembly of this invention can be used in conjunction
with a conventional illuminator, such as a fluorescent bulb. While
a CCD element is employed according to a preferred embodiment,
other types of electro-optical imaging components can be employed
in the camera assembly. The CCD utilized can be implemented as a
gray scale camera or as a color camera having three or more lines
of associated color pixels. The optical, signal processing,
illumination and resolution control techniques described herein can
be adapted to operate in conjunction with multiple lines of color
by those of ordinary skill. Similarly, the reflector arrangement
can be adapted to alternative camera arrangements. Additional
calibration procedures and filtering techniques can also be
employed. Such calibration techniques can be used to adjust the
performance and output of the camera element/CCD. Furthermore,
while not shown, it is contemplated, according to a preferred
embodiment that side guides can be provided to the feed tray the to
maintain sheets in a centered relationship within the feeding
mechanism. The guides can be adapted to move in conjunction toward
and away from each other along the widthwise direction using racks,
pinions and the like, based upon well-known arrangements.
Alternatively sheets can be registered against one of the widthwise
side edges of the tray. Accordingly, this description is meant to
be taken by way of example and not to otherwise limit the scope of
the invention.
* * * * *